EFFECTS OF RIDE MOTION PERTURBATION ON THE SPEED AND ACCURACY OF IN-VEHICLE REACHING TASKS by

Kevin Andrew Rider

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Industrial and Operations Engineering) in The University of Michigan 2006

Doctoral Committee: Professor Don B. Chaffin, Co-chair Associate Professor Bernard J. Martin, Co-chair Associate Professor Richard Brent Gillespie Associate Research Scientist Matthew P. Reed

© Kevin A. Rider All rights reserved 2006

Dedication To Loretta – my best friend, and my wife.

ii

Acknowledgements I am profoundly grateful for the professional and personal support and encouragement that I have received from my friends and colleagues over the past several years. With regard to the research presented in the following document, the mentorship provided by Don Chaffin, Matthew Reed, and Bernard Martin has helped shape my abilities to dream, develop, discern, and disseminate contributions in all areas of academic life. I will remain indebted to their sacrifice of time and effort to assist me in my endeavors. Their efforts that preceded my doctoral studies paved the way for my research within the Automotive Research Center (ARC), a research partnership between the University of Michigan, the US Army, and industrial corporations, with a seemingly endless list of people that have a genuine desire to assist others. This research would not have been possible without the support of the US Army – Research, Development, and Engineering Command (RDECOM), specifically the Motion Base Technologies team, including Harry Zywiol, Kyle Nebel, Victor Paul, and Annemarie Meldrum. The staff of the HUMOSIM Laboratory and the Center for Ergonomics, especially Charles Woolley, were instrumental in supporting the needs of my research. The additional support that I received from other graduate students also proved to be essential components in my research development: from early

iii

collaboration with Woojin Park and Clark Dickerson, to later accountability with David Wagner and Suzanne Hoffman. I wish all of these people the best, and many thanks. I also want to thank my wife, Loretta, and our children (Andrew, Ryan, Elizabeth, and Caleb) – while they may not always have helped my progress, they certainly helped me maintain my sanity and perspective. Loretta, through her continual encouragement and accountability, has proven to be an invaluable resource for me. This journey may not have happened, if not for her unequivocal support. I cannot repay her for the sacrifice that she has made in order to support me during these years, but I plan to spend the rest of my life trying.

iv

Table of Contents Dedication

ii

Acknowledgements

iii

List of Figures

vi

List of Tables

viii

Abstract

ix

Chapter 1 - Introduction

1

Chapter 2 - Background literature and experimental summary

10

Chapter 3 - Study 1: Investigation of the human biodynamic response to vertical sinusoidal vibration

26

Chapter 4 - Study 2: Analysis of the effects of vibration frequency, magnitude, and direction on movement times 48 Chapter 5 - Study 3: Effects of ride motion on movement time and accuracy

70

Chapter 6 - Use of visual and somatosensory feedbacks under ride motion

103

Chapter 7 - Modeling active human biodynamic response during reaching

127

Chapter 8 - Summary and future research opportunities

144

References

149

v

List of Figures Figure 2-1. Digital human model used to animate BIODYN-II predictions. 14 Figure 2-2. Illustration of how reach direction is incorporated into movement time predictions, using a sinusoidal multiplier. 22 Figure 3-1. Simplified human body sub-systems and primary vibration resonance bands. 28 Figure 3-2. Laboratory setup, including Ride Motion Simulator (RMS) cab, and cameras from the VICON motion capture system. 31 Figure 3-3. RMS with HMMWV instrument panel used in Study 1. 32 Figure 3-4. Digital human 3D environment with pushbutton locations indicated. 34 Figure 3-5. Digital representation of RMS cab with VICON camera setup indicated. 35 Figure 3-6. Jack digital human depicting reflective marker locations, and the subject in the home posture. 36 Figure 3-7. Mean fingertip variability at the target (an estimate of the effective target width), as a function of vibration frequency and target location. 40 Figure 3-8. Illustration of correlation of the frequency effects related to fingertip variability, subjective difficulty ratings, and relative increases in movement times. 42 Figure 3-9. Variability of effective target width under vertical frequency sweep. 43 Figure 4-1. Depiction of reaches to targets that did not require use of the torso. 53 Figure 4-2. Depiction of reaches to targets that required torso movements. 54 Figure 4-3. Participant seated in the RMS cab with hands in the home position on the steering wheel. 55 Figure 4-4. Digital human with reflective markers in the home resting posture. 56 Figure 4-5. Variability in movement times between participants for self-paced reaches in which subjects were only instructed to successfully push the target. 58 Figure 4-6. Lognormal distribution of movement times showing constant variance. 58 Figure 4-7. The interaction effect of vibration frequency and vibration magnitude on mean movement times. 61 Figure 4-8. Effect of vibration frequency and vibration direction on mean movement times. 63 Figure 4-9. Interaction vibration direction and target location on movement times, normalized by movement times in the stationary condition. 65 Figure 4-10. Deviations to the reach trajectory from a straight line are very similar, with exceptions of the Forward Near and Forward Far targets, to which reaches were required to avoid the steering wheel. 66

vi

Figure 5-1. RMS platform, reconfigurable cab, and experimental setup including three touch-screens. 75 Figure 5-2. A power spectral density of the angular acceleration commands input to the RMS motion platform for the pitch dominated drive file. 77 Figure 5-3. The RMS configured with three touchpanel displays. 80 Figure 5-4. Target configurations: a) Digital targets b) Physical targets. 81 Figure 5-5. Digital human with reflective markers in the home resting posture. 82 Figure 5-6. Experimental block design with overlapping conditions, including control factors in each block. 83 Figure 5-7. Increases in reaction time and effective target width due to ride motion (across all reaching tasks). 88 Figure 5-8. Effects of target size and target type on the timing and accuracy of reaching tasks. 93 Figure 5-9. Effect of ride motion on reach accuracy, as a function of the distance from the reach endpoint and the center of the target. 96 Figure 5-10. Interactions of ride motion, display location, and target type. 97 Figure 6-1. Speed profiles: a) Subsequent discrete optimal movements, b) Movement resulting from superposition of planar movements. 106 Figure 6-2. The RMS configured with three touch-screens. The participant is illustrated reaching to a target on the Lateral touch-screen. 112 Figure 6-3. a) Illustration of radial mapping used to identify deviations of fingertip position relative to the mean trajectory at the peak velocity (b) and the target (c). 116 Figure 6-4. Illustration of the mapping of the deviation of the fingertip position at peak velocity to the reach endpoint on the display. 116 Figure 6-5. Endpoints without visual feedback under stationary and ride motion conditions to each of the three touchscreens: a) Forward-Up, b) Forward, and c) Lateral. Dashed ellipses contain 95% of the reach endpoints for that condition. 118 Figure 6-6. Mapping of spatial dispersion of reach endpoints for each 120 display (columns) and motion condition (rows). Figure 7-1. Example of a dynamic kinetic model of a seated vehicle operator used to calculate the passive biodynamic responses to sagittal plane ride motion while holding a steering wheel. 132 Figure 7-3. Illustration of normally distributed “launch cone” incorporating the initial trajectory for 95% of empirical reaches to a target location. 137 Figure 7-4. Comparison of spatial dispersions of the fingertip at the peak velocity in a stationary cab (Top), and under ride motion (Bottom). 138 Figure 7-5. Adjustment to reference trajectory based on linear extrapolation 139 of the deviation at the peak velocity. Figure 7-6. Typical example of superposing empirical fingertip trajectories and vibration feedthrough accelerations. 140 Figure 7-7. Illustration of the Hermite curve using the tangential vectors of 142 current and desired velocity vectors. Figure 7-8. Architecture of STORM algorithm, and component integration. 143

vii

List of Tables Table 3-1. ANOVA results for the fingertip variability at the target during vertical sinusoidal vibrations. Table 4-2. Participant summary: Stature (cm), Weight (kg), Age (yr). Table 4-3. ANOVA results for lognormal movement times for distancenormalized reaches without temporal constraints. Table 4-4. ANOVA results for ratio of Frechet distance to reach distance. Table 5-5. Participant summary: Stature (cm), Weight (kg), Age (yr). Table 5-6. Acceleration data for 6DOF ride profile. Lateral, longitudinal, and vertical data are reported in m/sec2. Roll, pitch, and yaw data are reported in rad/sec 2. Table 5-7. Overlapping conditions of experimental design. Table 5-8. ANOVA results for reaction times from combined results. Table 5-9. ANOVA results for normalized movement times from combined results. Table 5-10. ANOVA results for radial distances from combined results. Table 6-11. Participant summary: Stature (cm), Weight (kg), Age (yr). Table 6-12. Summary of dependent measures with respect to ride motion and visual feedback. Table 6-13. ANOVA results for radial deviation of the fingertip at the target.

viii

39 52 60 67 75

76 83 89 92 95 111 119 121

Abstract EFFECTS OF RIDE MOTION PERTURBATION ON THE SPEED AND ACCURACY OF IN-VEHICLE REACHING TASKS by Kevin Andrew Rider

Co-chairs: Don B. Chaffin and Bernard J. Martin The ability to quickly and accurately perform precise manual reaching tasks while riding in moving environments is a well recognized problem. Low frequency vibrations, typical of vehicle ride motions result in biodynamic perturbations to the extended arm, inhibiting the quick and accurate reaching movements

that

are

necessary

to

operate

control,

navigation,

and

communication systems. Through this investigation, the effects of ride motions on movement time and endpoint accuracy have been systematically investigated with respect to the location and characteristics of targets, and vehicle ride motions. Vehicle motions were reproduced utilizing a Ride Motion Simulator, in which over 26,000 whole-body seated reaches were performed. Movements were recorded using a motion capture system for subsequent kinematics analyses. Ride motions are shown to significantly affect the speed and accuracy of reaching tasks, performed with and without temporal constraints. For temporally-

ix

unconstrained reaches, increases in vibration magnitude resulted in 6%, 10%, and 16.5% longer movement times for low, moderate, and rough vibrations, respectively. Compared to reaches performed in a stationary environment, reaches in directions that coincided with the direction of vibration resulted in 6.6% longer movement time, while reaches in directions perpendicular to the vibration direction were 11% longer in duration. For rapid reaches, whole-body ride motions contributed to approximately 13.5% longer reaction times, 3.6% shorter movement times, and a 56% increase in endpoint variability. Visual and proprioceptive feedbacks also were investigated, where endpoint variability was approximately 3.3 times larger without vision of the hand than visually-guided reaches, while ride motions contributed to an additional 30% increase in endpoint variability. Collectively, ride motion perturbations of visually-occluded reaches resulted in 4.5 times larger endpoint variability than visually-guided reaches performed in a stationary cab. An active biodynamic human response algorithm is proposed to simulate seated reach trajectories under ride motion perturbations, incorporating trajectory planning, vibration feedthrough to the hand, and visual and proprioceptive feedbacks. This investigation and model development quantifies the performance degradation of in-vehicle reaching tasks when experiencing typical off-road vehicle motions. Direct extrapolation of these results to other vehicles or ride environments is not recommended.

x

Chapter 1 Introduction This chapter outlines the problems associated with extended arm reaching movements in a dynamically unstable environment. The specific context is performing rapid in-vehicle reaching and pointing tasks while being subjected to vehicle ride motion.

Problem Description During the normal operation of moving vehicles, operators are regularly required to perform reaching and pointing tasks to manipulate controls and operate vehicle navigation and communication systems. The ability to activate pushbutton controls and touchpanel displays quickly and accurately can be severely impeded by the dynamic motion of the vehicle. While common experience shows that even writing and drinking are difficult in a moving vehicle (Corbridge and Griffin 1991), military and construction vehicles are prominent examples of scenarios that cannot afford significant disruptions to operators’ movements. The completion of target identification and acquisition tasks is essential to the success of mission-critical military scenarios, where ride motion perturbations may unacceptably alter the accuracy of movements of the unsupported arm. Construction equipment operators also are required to manipulate vehicle controls under vibration conditions. Although temporal task constraints may be relaxed compared to the military case, extreme lateral and 1

longitudinal movements of the cab and operator can severely degrade performance, and contribute to rollover accidents, which comprise nearly 20% of all construction vehicle deaths (EASHW 2006). The operation of on-road passenger and commercial vehicles also are significantly impacted by ride motion, where increases in reach difficulty due to vehicle motion translate into increased task times, which are correlated to driver distraction and degraded vehicle control (Tijerina 2000). Specifically, Wang et al. (1996) showed that movements associated with secondary tasks, such as tuning the radio, and operating controls, are some of the most common causes of inattention-related crashes. Furthermore, Wierwille and Tijerina (1996) estimated that 55% of inattention crashes result from interaction with objects, people, and instrumentation within the vehicle. Improving the layout and design of vehicle controls and displays may reduce the time required to perform in-vehicle tasks, and decrease associated safety concerns. Although the automobile industry has long been aware of the connection between vehicle control design, usability, and vehicle safety, improvement of interior design often remains a subjective task (Green 2000). More quantitative tools are needed to evaluate the expected performance associated with using proposed designs. Technological advances in controls and displays have resulted in an increased number of pushbutton controls and displays to be located in the vehicle cockpit, providing increased functionality at the operator’s fingertips (Zwahlen 1993). However, due to the finite amount of working space within the occupant’s functional reach envelope, controls are becoming smaller, and

2

located further from the driver. Conversely, in-vehicle touchscreens for navigation, communication, and controls can reduce the number of controls that are required through use of hierarchical menus of commands, but also results in increased task time due to the cognitive processing required navigating through the menu system (Aretz 1991). The difficulty of these respective tasks is often associated with the size of the target and the distance of the reach, as in Fitts’ law (Fitts 1954; Fitts and Peterson 1964), which is described in more detail in Chapter 2. The increased difficulty of performing these tasks in a dynamic environment is associated with the ride motions transmitted through the vehicle subsystems (i.e. tires, suspension, seating) and to the vehicle operator, which result in disruptions to the intended reaching movements. These perturbations result in degraded accuracy as the trajectory of the end effector, nominally the hand or fingertip, is perturbed away from its intended path. The operator is then required to discern the error and execute compensatory movements that will enable the successful completion of the reaching task (Bernstein 1967, Schmidt and Lee 2005). Degraded manual performance in enclosed motion environments has been documented in several research studies, where reduced cognitive and fine motor performance was observed during motion and persisted after motion stopped (Lewis and Griffin 1979; Martin et al. 1981; Cowings et al. 1999). The vibration magnitude and duration of the exposure have also been shown to influence cognitive performance (Schipani et al. 1998). The performance degradation observed in moving vehicles is further supported by laboratory

3

studies, which have demonstrated that a motion environment can influence other specific aspects of task performance. Examples include 1) vibration can degrade sensory inputs (Goodwin et al. 1972; Roll and Vedel 1982; Gauthier et al. 1981; Ghez 1991; Ghez et al. 1990), 2) physical perturbations and visual distortions can degrade movement control (Polit & Bizzi 1978; Contreras-Vidal & Kerick 2004), and 3) postural instability can change central nervous system processing (Bittner and Guignard 1985; Redfern et al. 2002). However, understanding the particular conditions under which motion environments degrade performance must be more thoroughly investigated. That is, the mechanism of the degradation will depend on the specifics of both the motion environment and of the task being performed (see Griffin 1990 for review). Some difficulties of completing the manual task in the dynamic environment are the perturbations that alter planned movements, and the potential mismatches between the occupants’ sensory systems. Motor planning and control theories suggest that both of these factors can lead to degradations in reach performance (Erlhagen and Schöner 2002; Wolpert and Ghahramani 2000; Wolpert et al. 1995). The decreased ability to predict environmental conditions, combined with a sensory mismatch, is likely to require additional time for movement planning and degraded manual control. The extent to which the operator’s reach performance is affected by ride motion may be a function of both the dynamic ride motion environment and the reach task requirements, though the individual and interactive effects are not well understood. It is then

4

necessary to evaluate the effects of terrain-induced ride motion with respect to the expected performance (i.e. speed and accuracy) of in-vehicle reaching tasks.

Objective The research reported in this dissertation attempts to extend the existing knowledge describing the human responses to vehicle vibration and associated human movement control. It is proposed that such knowledge will provide a foundational understanding of the specific ride motion effects that contribute to degraded manual performance in a vehicle. It further proposes an active human biodynamic response model that can simulate human reach trajectories to targets within the vehicle when being perturbed by moderate vehicle ride motion.

Thesis organization The nature of this research is to understand and predict the effects of six degree of freedom (6DOF) ride motion on the speed and accuracy of in-vehicle reaching tasks. It will be shown that the ability to perform these tasks is a function of the dynamic environment and the task parameters. This interaction will be explored through the investigation of the biodynamic response of the seated operator to vehicle motion and the capability of the neuromuscular system to plan and execute successful reaching movements. This understanding will provide a robust foundation for related and future research on the human-centered design of controls and displays within the cockpit of a moving vehicle. Chapter 2 provides the necessary background on whole-body vibration and previous research on the biodynamic effects on the seated human operator, as well as a review of current movement theories and how they might apply to

5

the problem at hand. The studies reported in the subsequent chapters investigated the contribution of several factors, including target size and location, and the direction, frequency, and magnitude of ride motion, attempting to quantify the magnitude of their effects and interactions. An initial study is described in Chapter 3, exploring the use of a humanrated, ride motion platform to simulate vehicle ride motions, and an optical motion capture system to record the movements of the seated operator performing invehicle reaching and pointing tasks. This study explored the effects of vertical sinusoidal vibration on reach accuracy, including subjective difficulty ratings that correlate reach difficulty to endpoint variability. The study effectively combines the largely dissociated knowledge bases of human vibration and human motor control by measuring the speed and accuracy of pointing movements while being subjected to ride motion perturbations. The effective target width is calculated, and an interaction between vertical vibration and reach direction is explored. Vertical frequency sweeps are utilized to reveal the biodynamic response of the extended arm and to study the frequency response of the maintained posture. Based on these initial findings pertaining to the frequency-dependent reach endpoint variability, a systematic factorial analysis detailed in Chapter 4 was conducted to quantify the principal main effects and interactions of environmental factors (i.e. vibration direction, vibration frequency, and vibration magnitude) and task parameters (reach direction and reach distance) on movement time. Each of these factors is shown to significantly affect movement

6

times, and interesting interactions include the vibration direction with reach direction, and vibration direction with vibration frequency. A final comprehensive study is described in Chapters 5 and 6, where movement times and endpoint accuracy of rapid pointing tasks were investigated with respect to target size and target type. Resistive touchpanel displays were used to record finger contact locations at the target, from which the spatial endpoint dispersions were calculated and analyzed. Chapter 5 summarizes the factorial analyses of the independent factors and their interactions on the dependent measures: reaction time, movement time, and endpoint accuracy. The speed-accuracy tradeoff is confirmed with respect to increased endpoint variability correlating to shorter movement times. However, ride motion perturbations result in both increased movement times and increased endpoint variability, suggesting that movement time predictions must also account for the dynamic environment in addition to the task parameters, such as movement amplitude and target width. The effects of ride motion on feedback-based movement control are presented in Chapter 6, where visually-occluded reaches were studied to investigate the independent effects of vision and proprioception on the performance of the intended reaching tasks. Finally, Chapter 7 presents an algorithm to simulate in-vehicle reach trajectories, integrating empirically-modeled movement variability, passive vehicle and biodynamic responses, and feedback control mechanisms. Separate models are discussed for the movement planning, movement generation, ride motion perturbations, and feedback controllers that are used to generate and

7

modulate reach trajectories in a dynamic environment. Movements are planned based on the initial location of the hand and the desired endpoint of the reach. Hand trajectories have been shown to be largely consistent across trials for the same subject in highly-learned movements (Soechting and Lacquaniti 1981), thus functional regression techniques were employed to generate reach trajectories, based on extensive empirical data collection (Faraway 2000, 2004). Natural variability has been observed that induces errors in the movement execution (Schmidt and Lee 2005), and this variability is modeled using a normally distributed cone of departure vectors that create a small deviation of the intended trajectory. Using a passive biodynamic human response model, described in Chapter 3, ride motion perturbations are simulated, and the resulting perturbation of the intended movement is calculated (Liang et al. 2005). This perturbation then disturbs the ongoing planned trajectory creating deviations that are detected by a feedback system. To ensure that the initial 100 ms are purely open loop, the feedback controller remains offline until 100 ms of movement have elapsed. At this point, proprioceptive information allows for minor trajectory compensations in the cases that the trajectory deviation exceeds a minimum threshold. Also after 100 ms from movement onset, the visual feedback controller begins making corrections when the fingertip enters the field-of-view, based on previous assumptions of slow (peripheral) and fast (central) visual feedback loops working together to make rapid trajectory corrections (Blouin et al. 1993). Lastly, the feedback controller employs a Hermite spline curve that connects the current

8

position and velocity to the desired endpoint (Faraway 2004). The biodynamic and feedback control systems continuously operate until the fingertip successfully hits the target. The trajectory that results from the feedback-controlled response to ride motion perturbations contains two important design features: the movement time required to complete this task and the location of the fingertip at target contact. Monte Carlo simulations can be run using this methodology to determine distributions for both movement time and spatial accuracy. Furthermore, temporal constraints can be varied to evaluate the effect on the spatial dispersion of endpoints, or vice versa, where spatial constraints can be varied to evaluate the effects on movement time. The robust nature of the model allows additional parameters to be adjusted for further evaluation of the factors related to manual control performance under ride motion. One example might include varying the orientation and/or distribution of the departure vector (i.e. movement variability), providing insight into the possible performance of a novice operator, as compared to a highly-trained operator. Finally, the overall ability to modulate the parameters of the speed-accuracy tradeoff in a dynamic environment provides a unique mechanism by which to evaluate the size and location of the controls and displays of vehicle cockpits.

9

Chapter 2 Background literature and experimental summary This chapter describes the principal factors that relate to the performance of in-vehicle reaching and pointing tasks under the perturbation of whole-body vehicle ride motion. In order to reduce confusion with referenced human vibration studies, “vibration” will be used to reference single-axis sinusoidal motions, and “ride motion” will refer to the six-degree-of-freedom motion of the platform-sitter interface. Vibration-related research is largely divided into two categories: wholebody vibration (WBV) and hand-arm vibration (HAV). The present work predominantly investigates the former, and considers the latter only where it specifically contributes to the observed ride motion effects.

Human Response to Vibration Low frequency vehicle motions generally originate at the ground-tire coupling and are transmitted through the tires, suspension, and seating before reaching the seated operator. The ride motions experienced by the vehicle occupants induce primarily a biodynamic response of the human’s musculoskeletal systems. In cases such as rough and off-road driving scenarios, ride motions can have more deleterious effects that include fatigue, discomfort, and even low back pain; although clear dose-response relationships have been difficult to obtain (Griffin 1990; Bovenzi and Hulshof 1998; Pope et al. 1998; Lings and LeboeufYde 2000). Discomfort and fatigue exposure limits have been defined by the 10

International Organization for Standardization (ISO 2631:1997) though military, construction, and forestry vehicles regularly traverse rough terrains that can generate seat accelerations exceeding these boundaries. The biodynamic response mechanisms of the seated body are complex and are not well understood or even identified, although some mechanical principles are clearly at work. The biomechanics of the human body under vibration has been modeled using masses, springs, and dashpots, whereby the transmitted vibration can be amplified or attenuated through each segment within the body (Coermann 1962; Amirouche and Ider 1988; Rosen and Arcan 2003). The transmissibility of vibration through the body is often measured at two points in the body to define the transfer function that relates the input signal to output acceleration signals. The combined transfer functions of the body segments can induce unintended movements of the hand, known as vibration feedthrough (also as “breakthrough” or “vibration-correlated error”), which impedes the ability to perform precision manual tasks. The postures associated with reaching tasks (i.e. the uncoupled or cantilevered arm) make the arm particularly susceptible to vibration feedthrough. Intermediate or distal resting places for the elbow, wrist, or hand can greatly improve task performance, though these are not always available or desirable, due to the constrained workspace within vehicle cockpits. With respect to the biodynamic response at the hand, frequencies input to the seat higher than 20 Hz are attenuated, while low frequencies tend to be amplified – most significantly in the frequency ranges that cause resonance

11

within the mechanical linkage of the torso (Kitazaki and Griffin 1998). Coermann (1962) identified resonant effects of the seated torso by revealing differences above 2 Hz in the dynamic response of seated individuals compared to a rigid mass. Similar findings were reported for both lateral and fore-aft vibrations in the 1-2 Hz range (Whitman and Griffin 1978). Hagena et al. (1985) and Sandover (1983) suggest that these vibrations near the principal resonance cause a pitching motion of the pelvis as a result of the seat-buttock coupling, inducing bending motions of the lumbar spine (Wilder et al. 1982; Pope et al. 1991). These pelvic and lumbar movements then result in movements of the upper torso, head, and arms, resulting in the observed low frequency vibration feedthrough that affects manual control performance. Increases in vibration magnitude have been shown to further degrade manual control performance (Miwa 1968; Lewis et al. 1978). However the relationship is nonlinear and interacts significantly with vibration frequency (Mansfield 2005). Ride motion magnitude can be directly related to the nature of the terrain, where terrains are often subjectively qualified as “smooth” or “rough”, depending on the subjective comfort of the vibrations experienced by the operator. The ISO 2631 standard suggests comfort limits that also depend on the vibration frequency and the duration of exposure. Johanning et al. (1991) detail nominal root-mean-squared (r.m.s.) vibration magnitudes ranging as low as 0.8 m/s2 r.m.s. in rail transportation, with transient shocks exceeding 1.5 m/s2. Maksimovic (1987) reported values more than 3.0 m/s2 r.m.s. in military tanks. Fothergill and Griffin (1977) add that “very smooth” roads typically produce

12

weighted acceleration magnitudes around 0.2 m/s2 r.m.s., while “moderate” roads produce magnitudes near 1.0 m/s2 r.m.s. Off-road vehicles, such as earth-moving and construction equipment may generate motions that exceed 2.0 m/s2 r.m.s. in “very rough” conditions. Although the effects of resonance associated with vibration frequencies, and their interactions with vibration magnitude are complex, passive human biodynamics models have been developed since the 1960’s, these models simulate responses to low frequency vertical vibration (Coermann 1962, Vogt et al. 1968). However simulation models of the three-dimensional human response to vibration is computationally intensive and their development has been impeded until recent technological and computational enhancements (Zhang 2001). One such model is BIODYN-II, a human biodynamic response model component within the AVB-DYN software package developed by Systems Technology, Inc. (STI, Hawthorne, California), which uses estimated neuromuscular parameters to calculate the dynamic response to vibration of the seated vehicle occupant. Planar models, originally developed in the 1970’s for the US Air Force, have been integrated into the present formulation, which includes a multi-body dynamics solver (Jex et al. 1978; Allen et al. 2001). Figure 2-1 illustrates a three-dimensional biomechanical model, Jack, which was used to animate the BIODYN-II biodynamic predictions (UGS 2006).

13

Figure 2-1. Digital human model used to animate BIODYN-II predictions. Illustration depicts relevant masses, and couplings to the seatpan, seatback, steering wheel, and floor.

STI describes their planar models in the following manner: “BIODYN-80 [and an equivalent lateral vibration biodynamic model] is a versatile computational tool used to determine transfer functions between vertical and/or fore-aft vibrational inputs and important biodynamic outputs, such as motions of the torso, head, eyes, arms, or hands. The program scenario assumes a seated operator, gripping an arbitrary-angle [wheel] and viewing a display, possibly engaged in a tracking task. The physical model uses an “isomorphic,” lumped parameter approach to represent the relevant portions of the whole-body torso, limbs, and head, as well as postural compliances among the joints. The implementation of this model includes a chain of interacting parallel and serial second-order elements, with neuromuscular and other force feedbacks at the arm or head. The resulting equations are in “second-order element” matrix form and apply to a wide range of seated postures… A variety of outputs and inputs can be specified to evaluate the desired transmissibility transfer functions” (Liang et al. 2005).

14

The formulation of many passive human responses models, including BIODYN, stipulate fixed joint stiffness parameters. These models are intended to simulate the biodynamic response associated with maintained postures, but may not be sufficient to predict the effects associated with dynamic reaching tasks. For example, regardless of the nature of the dynamic system in which they function (i.e. zero-, first-, or second-order), McRuer et al. (1965) suggest that people modulate the joint stiffness parameters that are relevant to the intended task in order to maintain a nearly consistent (i.e. stable) system output. These changes in joint stiffness affect the transmissibility of vibration, and the biodynamic response (Hogan 1984; Milner and Cloutier 1993, 1998). Administrative controls also can be incorporated to reduce the adverse ride motion effects in the following areas: at the source, during the transmission, or at the area that contacts the person. In ground-based vehicles, this generally translates to the ground-tire and engine-frame coupling, the vehicle suspension, and occupant seating, respectively. Manufacturers of each of these components can improve designs to attenuate ride motions; however any mitigation strategies must be a systems-level approach. Redesigning individual parts without consideration of the entire system can be myopic, resulting in undesirable resonances between subsystems, which may even amplify the vibration transmission. Ride motions, typical of rough-road vehicle motion, have been consistently shown to reduce a person’s ability to accurately perform continuous manual tracking tasks (Lewis et al. 1976, 1979; Gauthier et al. 1981; Corbridge and

15

Griffin 1991), however performance levels varied between these studies, due to the numerous factors that contribute to the manual performance degradation. Specifically it has been shown that the deleterious performance effects (Gauthier et al. 1981) and altered neuromuscular responses (Martin et al. 1984) persist after vibration has ended. Though the causes of deleterious effects of vibration are complex, psychophysical, biomechanical, and neuromuscular factors were investigated in the course of this research and are discussed in the following chapters.

Human Movement Control Within vehicle cockpits, reaching tasks are affected by vibration feedthrough, requiring operators to make compensatory adjustments to their movements in order to successfully complete the intended tasks. Particularly in military scenarios, this performance degradation may be unacceptable, even placing individuals in life-threatening situations if critical tasks cannot be completed both quickly and accurately. The effects of these ride motions on reach performance must then be more clearly understood, so that vehicle subsystems can be designed so occupants have the capability to perform any necessary precision reaching tasks. The planning and execution of human movements have been extensively studied, beginning more than a century ago, when Woodworth (1899) was one of the first to suggest an impulse-timing relationship for discrete pointing tasks. A primary contribution that has been widely accepted and largely maintained today is that motions can be generally segmented into two sequential phases: current

16

control, and adjustment. The current control is also referred to across disciplines as an open loop or feedforward movement, where an impulse force is generated by the neuromuscular system. The trajectory is then modulated in the adjustment phase through any necessary corrections to ensure the successful completion of the intended task. These corrections are commonly referred to as closed loop or feedback based movements. Bernstein (1967), Schmidt (2005), and many others have provided significant support in comprehensive summaries on movement control theories. Open loop activities are planned prior to movement, and executed without correction from any feedback received during the movement. An anecdotal example was given by Lashley (1951) where he showed that finger movements of expert pianists do not utilize feedback to ensure that the movements are performed correctly, and that the subsequent movements are not related to the performance of the previous movement. These open loop motions are almost universally very rapid, although some arguments can be made for longer sequences of highly-trained motions, as might be observed in skilled athletes throwing a ball or swinging a golf club (Keele 1968; Sternad and Schaal 1999). Conversely, closed loop activities integrate feedback from internal sensory systems and the environment, to evaluate performance and determine if corrections should be made to reduce any errors that are detected during the motion. Common examples include catching a ball and driving a car, both of which require processing and responding to continuous information to complete or maintain the intended task performance.

17

Two of the primary feedback mechanisms are vision and proprioception. With respect to vision, Keele and Posner (1968) estimated that the neuromuscular system could execute a movement based on visual information within 190-260 ms. Carlton (1981, 1992) also has provided strong evidence of visuomotor processing times around 135 ms by systematically adjusting the viewing and occlusion time of target stimuli. However, any estimates of neuromuscular feedback loops depend on which muscles are involved. Zelaznik’s (1983) estimates of reaching movements are more relevant to the present study, and are as short as 100 ms when the content of the viewed images was known a priori. Proprioception estimates the changing position of body segments within us, although the proprioceptive feedback (propriomotor) loop has not been adequately estimated (Ingram 2000). Specifically, sensory receptors that include the muscle spindles and Golgi tendon organs monitor the changes in muscle length and tension, respectively. By resolving these changes, the neuromuscular system can effectively monitor the desired manual performance and execute compensatory movements when deviations are detected away from the intended manual or postural control (Keele 1968; Prablanc et al. 1986; Flash et al. 1992; Flanagan 1993; Sabes 2000). The path of the reach trajectory has been shown to be linear in rapid, twodimensional, unobstructed reaching movements, for which the speed of the hand exhibits a bell-shaped profile (Morasso 1981, 1983). Studies investigating the interactions of reach direction and distance have been reported to substantiate this claim (Buck 1982; Gordon et al. 1994; Gréa 2000). The speed of the hand

18

has also been inversely correlated to endpoint accuracy, which is commonly referred to as the “speed-accuracy tradeoff” (Fitts 1954; Howarth et al. 1971; Abend et al. 1982; Flanagan 1993; Adamovich et al. 2001). The inverse relationship between speed and accuracy has been originally modeled by Fitts (1954), predicting movement time (MT) from the movement amplitude (A) and target width (W), as given in Equation 2-1 (Fitts 1954, Peterson and Fitts 1964). Fitts’ law utilizes regression coefficients to establish a log-linear equation for predicting movement time based on the Index of Difficulty (ID) of the task. This relationship has successfully predicted the movement times for widely-varying combinations of static environments, including movements with the arms and legs (Langolf et al. 1976), head movements (Jagacinski & Monk 1985), and even underwater (Kerr 1975).

⎛ 2A ⎞ MT = a + b × log2 ⎜ ⎟ ⎝W⎠

(2-1)

ID Fitts also showed that this relationship is applicable to single-aiming tasks of point-to-point discrete movements (Fitts & Peterson 1964). However more recent formulations, such as the Shannon formulation (Equation 4-2) proposed by MacKenzie (1992), have been shown to have stronger theoretical constructs and a better fit to empirical data (MacKenzie 1995).

⎛A ⎞ + 1⎟ MT = a +b × log 2 ⎜ ⎝W ⎠

(2-2)

Both of these formulations imply a stationary environment and a stationary target, neither of which may be the case during in-vehicle tasks. As a vehicle

19

operator attempts to reach and activate pushbutton controls or displays, ride motions cause unpredictable perturbations to the intended hand trajectory. These disturbances can cause deviations that require adjustments to the trajectory that often require additional time to complete the task (Meyer et al. 1988). If the target size and reach distance are held constant, then the movement times can be directly compared between stationary and moving environments to estimate the effect of ride motion on movement time and reach difficulty. It remains that the original and modified formulations of Fitts’ law are inherently one dimensional, with no accommodation for off movement errors perpendicular to movement direction, referred to as off-axis error (Gordon et al. 1994). Mackenzie and Buxton (1992) attempted to provide a mechanism by which to account for off-axis error by testing several possible combinations of target dimensions. However, their primary recommendation – using the smaller of the two dimensions – only reduces the 2D problem back to one dimension by choosing to simply ignore part of the dimensionality problem. More relevant to our problem, Murata and Iwase (2001) attempted to extend Fitts’ law to predict movement times for rapid three-dimensional pointing tasks in a stationary environment (Equation 4-3), with arguably more success.

⎤ ⎡ ⎞ ⎛A MT = a +b × ⎢log 2 ⎜ + 1⎟ + c × sinθ ⎥ ⎠ ⎝W ⎦ ⎣

(4-3)

Reach direction is incorporated in the Murata formulation as an additive factor of reach elevation in formulation of the ID, multiplied by an empirically fit constant, c. This multiplier nicely adjusts movement time predictions based on the relative difference in height between the target location and the origin of the 20

reach, without making adjustments for lateral displacements. The authors measured movement amplitude (A) as the distance between the hand and the vertically oriented target board, and thus do not accurately account for the actual movement amplitudes to the targets, which were located about the perimeter of the board (Figure 2-2). While their empirical fit has a high correlation, supporting a robust formulation for predicting movement times, movement amplitude must be redefined as the actual distance to the endpoint of the reach. To their credit, this directional factor affects the reach difficulty (ID), without being directly associated with movement amplitude. However reach direction seems unnecessarily removed from the logarithmic function, resulting in different treatments of horizontal amplitude and vertical elevation. The robust nature of the original formulation may mask this discrepancy or perhaps the constant provides compensations within the empirical fit. A more appropriate formulation might incorporate the amplitude and elevation of the reach into the numerator of the ID.

21

Upper left: 135°

Upper: 90°

Upper right: 45°

Center

Left: 180°

Lower left: 225°

Right: 0°

Lower: 270°

Lower right: 315°

Figure 2-2. Illustration of how reach direction is incorporated into movement time predictions, using a sinusoidal multiplier (adapted from Murata and Iwase 2001).

Schmidt et al. (1978, 1979) describe the relative spatial error of a reach as the effective target width (ETW), or the “within-subject variability in movement distance, calculated within a ‘nominal distance-movement time’ goal condition”. Fitts’ law directly relates the size of the “effective target”, or fingertip excursion, at the destination of the reach, to the difficulty of the task, and this relationship may also apply in moving environments, as shown in pilot studies described in Chapter 3. Soechting et al. (1983), Gréa et al. (2000), and Sabes (2000) provide arguments that Fitts’ law adequately predicts reach movement times even when the target is moving, unlike the stationary targets in the Fitts’ paradigm. In the case of in-vehicle reaching tasks, ride motion alters the position of the arm relative to the target, causing a visual perception of a moving target with respect

22

to the relative locations of the eyes and fingertip (Lee 1976; Jeannerod 1984; Gentilucci et al. 1992). A test of Fitts' law regarding moving targets (Jagacinski et al. 1980) agreed that the Fitts’ relationship held up “fairly well” in some dynamic situations, but did not find a strong predictive relationship. Several modifications and additions to Fitts’ Law have been proposed to better account for the variability of movement in various dynamic scenarios. These revised equations typically include an additional parameter or adjustment to the ID (Mackenzie and Buxton 1992) to account for ride motion perturbations. Many studies make a distinction between temporally and spatially constrained reaching tasks, where linear and logarithmic functions should be used respectively (Meyer et al. 1988, 1990). Speed profiles of the fingertip during rapid, “aimed-reaching tasks” are smooth and bell shaped, but the smoothness of the distribution degrades as temporal constraints are relaxed (Morasso 1981, 1983; Marteniuk 1987). Contrary to the impulse-timing hypothesis suggested by Woodworth, a more mechanical approach has been suggested; the neuromuscular system alters the joint stiffness parameters of an effective spring and dashpot structure resulting in coordinated joint movements. In this theory, agonist and antagonist muscular systems are activated through stretch reflexes until the system reaches an equilibrium stationary state. Asatryan and Feldman (1965) most notably began this reasoning, realizing that the relationship between a muscle’s length and tension could act like an elongated spring. Utilizing a spring’s linear length-tension relationship, Feldman (1966, 1986) offered that competing muscular systems acting as springs would find an equilibrium position about a single pin joint, where

23

changes to the effective length of either “spring” would move the overall system toward a new equilibrium point (E-P). This theory has since been extended to multi-joint, unrestrained arm movements that are regulated by the joint position and endpoint of the limb (Mussa Ivaldi et al. 1985; Latash 1993; Latash et al. 1999), and later whole-body movements (Domen et al. 1999). Polit and Bizzi (1979) provided further support of the E-P hypothesis through studies with rhesus monkeys that performed visually-occluded arm pointing tasks before and after surgery that removed proprioceptive feedback. Preoperative and postoperative movement analysis showed similar endpoint forearm positions despite the lack of feedback, suggesting that nonvolitional corrections result from a determination of the joint stiffness parameters prior to movement onset. Similarly in humans, Schmidt (1980) showed that unperceived displacements in limb position also were quickly compensated. However a latency observed before the response in both studies showed that the systems are not purely mechanical (Dewhurst 1967). A combination of the impulse-timing and E-P hypotheses was suggested by Berkenblit et al. (1986), in which a coordinated series of equilibrium points could generate movements. Plamondon (1995a, 1995b, 1997) developed a deltalognormal law, which incorporates seven parameters of the agonist and antagonist systems in predicting two-dimensional movement (thirteen parameters for three-dimensional motions), and is based on the same serial E-P theory. Further conjecture of these models toward effective movement control strategies are discussed in Chapter 6.

24

Though numerous studies have analyzed the individual factors that affect reaching motions (Flash et al. 1992; Messier and Kalaska 1997; Adamovich 2001), the complex interaction of ride motion conditions and target location that affect the movement times and endpoint accuracy of reaching remains largely unknown. By systematically evaluating the effects of ride motions on specific reaching tasks, the principal factors affecting reach performance can be identified.

25

Chapter 3 Study 1: Investigation of the human biodynamic response to vertical sinusoidal vibration This study was intended to investigate the extent to which sinusoidal vertical vibration degrades 1) the ability to maintain a stationary posture in a dynamic environment, and 2) the endpoint accuracy of in-vehicle reaching tasks.

Background The performance of manual control tasks is known to be negatively affected by whole-body vibration (WBV) (Lewis and Griffin 1978; Gauthier et al. 1981; Martin et al. 1991), and this performance degradation can have significant effects on a vehicle operator’s ability to quickly and accurately perform in-vehicle reaching tasks, such as in the operation of many advanced vehicle control, navigation and communications systems (Zwahlen 1993). The terrain-induced vehicle vibration, hereafter known as ride motion, is transmitted through the vehicle subsystems (e.g. tires, suspension, and seating) to the seated vehicle operator, where this vibration feedthrough can cause significant disturbances to the operator’s posture (Kitazaki & Griffin 1998), particularly in frequency bands near the natural resonant frequency of the torso (Matsumoto & Griffin 1998; Matsumoto & Griffin 2001). This postural effect modulates the transmission of vibration through the body (Messenger et al. 1989; Paddan & Griffin 1993; Mansfield & Griffin 2000; Mansfield & Griffin 2002). Ride motion increases the

26

co-contractive muscular forces used to stabilize the torso, and contributes to the changes observed in the joint stiffness and damping parameters of the torso and extended arm (Milner 2002; Wakeling et al. 2002), which affects the mechanical impedance of the torso (Coermann 1962). Since in-vehicle reaching tasks typically require bending or twisting the torso, the postural effect on reach performance must be further evaluated with respect to extending an unrestrained arm in a dynamic environment. Frequency sweeps, or chirp signals, can be used to identify resonant frequencies within a body by inducing sinusoidal vibration with continuously increasing frequency within a frequency band of interest. It is known that segments of the human body are particularly sensitive to vibration between 0.5 Hz and 30 Hz. In seated reaching movements, vibration is transmitted through the buttock, spinal column and abdominal mass, shoulder girdle, upper arm, lower arm, and hand, where the resonance bands for these segments range from 4 Hz to 30 Hz (Figure 3-1). Iso-comfort curves have been developed to quantify the tolerance of the seated vehicle occupants to dynamic ride motion, particularly in this frequency range (e.g. ISO 2631-1:1997). These curves indicate limits of the “perceived” discomfort, as a function of vibration frequency and magnitude. Leatherwood et al. (1980) reported an increase in sensitivity to vertical vibrations between 4 and 8 Hz, consistent with the peak resonance of the torso in seated humans. Subjecting a seated vehicle occupant to a vertical frequency sweep in these ranges should excite the resonance of the aforementioned body segments;

27

hence, the effects on manual control and coordination can be explored. The resulting movement of the fingertip due to the vibration feedthrough can then be measured and used as a metric of reach difficulty and/or ride comfort.

Figure 3-1. Simplified human body sub-systems and primary vibration resonance bands (SafetyLine Institute 2006).

Due to the complex biodynamic response of the human body, vibration research has commonly focused on quantifying the effects of single-axis vibration. In general much of the manual control studies have evaluated the biodynamic effects of random ride motion on vehicle occupants whose hands are rigidly coupled to a maneuvering device, such as a control stick or steering wheel (Lewis

28

and Griffin 1976; Gauthier et al. 1981, Martin et al. 1991), though a few studies pertain to reaching movements in dynamic environments (e.g. Martin et al. 1984). On the other hand, the “motor control” community has investigated the dexterity of human reaching for over a century and has developed models that can be used to predict reach trajectories and endpoints even in the presence of a disturbance to the movement. However these perturbations are typically either a single impulse or constant force field, and are rarely a random variable. Examples of these models are discussed further in Chapter 6. The vibration and motor control literature support the position that terrain-induced ride motion will degrade manual reach performance in general; however the magnitude of this degradation under varied ride motion conditions has not been investigated. In the present study the following hypotheses were specifically tested to determine the conditions that seem to most significantly degrade reach performance during ride motion: Hypothesis 3.1: The effective target width of point-to-point reaches under whole-body vertical vibration is significantly greater at frequencies between 4 Hz and 6 Hz, Hypothesis 3.2: Target location significantly affects the effective target width of point-to-point reaches under vertical whole-body vibration, Hypothesis 3.3: Perceived reach difficulty is correlated with the effective target width, and Hypothesis 3.4: Fingertip variability of maintained reaching postures during the vertical frequency sweep is greater when the arm is fully extended than in postures where the hand is closer to the body.

29

Materials and Methods PARTICIPANTS Two participants volunteered for this study, one man (172.2 cm, 77.3 kg) and one woman (167.6 cm, 59.1 kg). They were both briefed about the purpose of the study and experimental procedures, and completed a human use consent form as part of both the US Army and University Institutional Review Board procedures. RIDE MOTION SIMULATOR (RMS) This study was conducted using the human-rated Ride Motion Simulator (RMS) at the U.S. Army Research, Development, and Engineering Command in Warren, Michigan. The RMS is a servo-hydraulic, hexapod simulator with six degrees of freedom, and is certified for human-in-the-loop studies (Figure 3-2). Some relevant technical specifications that describe the motion platform are provided below: • • • • •

Payload: 1,600 lbs Platform diameter: 46 in Axes: linear maximum displacement: ± 20 in Axes: linear peak velocity: ± 50 in/s Axes: linear peak acceleration: ± 2 g’s

30

Figure 3-2. Laboratory setup, including Ride Motion Simulator (RMS) cab, and cameras from the VICON motion capture system.

A trained RMS operator verified the operational status and safety of the system prior to testing by sequencing through all of the combinations of input frequencies and amplitudes of the experiment. Safety limits were engaged and emergency-stop buttons were available to the experimenter, the participants, and the spotter, who visually monitored the physical response of the subject during the test sessions. The RMS cab was instrumented with a High Mobility Multipurpose Wheeled Vehicle (HMMWV) instrument panel, shown in Figure 3-3.

31

Figure 3-3. RMS with HMMWV instrument panel used in Study 1.

The RMS simulated 0.5 g and 0.8 g peak-to-peak (p-p) sinusoidal inputs at the seat-sitter interface over a range of frequencies, while two seated subjects performed reaching tasks to six targets located in common vehicle control locations. Additionally, the subjects held terminal reach postures during a singlepass, logarithmic frequency sweep from 0.5 Hz to 32 Hz. The frequency doubled every twelve seconds at a constant acceleration of 0.5 g p-p (1.734 m/s2). Reach kinematics were recorded using a 10-camera VICON motion capture system. The effects of vertical ride motion on movement time, accuracy, and subjective responses were assessed.

32

Because past studies of vibration have largely determined that frequencies between 4 and 6 Hz, also known as a principal human resonant frequency band, most significantly affect the performance of manual tasks, these and contiguous lower and higher frequency bands were used. These included the sinusoidal vertical frequencies of 2, 4, 5, 6, 8, and 10 Hz, with vertical accelerations of 0.5 g and 0.8 g p-p applied to the motion platform at each frequency. REACH TARGET LOCATIONS Six circular pushbuttons with diameters of 12.8 mm were placed in the participant’s right-hand reach envelope. Nominal names were assigned for approximate in-vehicle locations; values in parentheses denote the approximate straight-line distance from the predicted location of the hip joint, or H-point, in the sagittal plane to the target. Figure 3-4 shows the EDS PLM Solutions’ Jack figure in a digital mockup of the RMS cab and six targets (indicated by the dark button icons) in the following locations: 1. “Overhead” - Above right shoulder (~1.2 m) 2. “Speedometer” - Forward of right shoulder (~1.0 m) 3. “Steering Column” - Forward of elbow, arm at rest (~0.5 m) 4. “Radio” - Elbow height, arm at rest, 45° right (~0.8 m) 5. “Glove box” - H-point height, 45° right (~1.0 m) 6. “Floor” – on floor, 45° right (~0.6 m)

33

Figure 3-4. Digital human 3D environment with pushbutton locations indicated.

MOTION CAPTURE Movements of the RMS cab and the subjects were recorded by a VICON 524 motion capture system, sampling at 60 Hz using ten cameras. Eight standard-sized analog cameras and two small “lipstick” cameras with 640 x 480 pixel resolutions were positioned around the rear of the semi-enclosed cab. Of those cameras, six were placed in overhead positions and the remaining four cameras were placed behind the cab, two on each side as shown in Figure 3-5.

34

Figure 3-5. Digital representation of RMS cab with VICON camera setup indicated.

Six reflective markers were placed at extreme locations on the RMS cab to record its motion. A headband with four markers, and one marker affixed to the skin on top of the bony protrusion of the C7 vertebrae, were used to collect data on the position and orientation of the head during the reaches. Elastic bands with attached markers were worn on the elbow, wrists, knees, and feet. Data from these markers were used to calculate the orientation of the arm and legs. In total, twenty reflective markers were also placed on the participant’s body, as shown in Figure 3-6.

35

Figure 3-6. Jack digital human depicting reflective marker locations, and the subject in the home posture.

EXPERIMENTAL PROCEDURE Reach to targets study. Participants reached to the six target locations under constant acceleration of 0.8 g (p-p). In addition to a “no-motion” condition, input frequencies of 2, 4, 5, 6, 8, and 10 Hz were applied in a random order. Participants performed two sets of reaches to each of the six target locations, and were given instructions to hold the outstretched index finger on the target for two seconds before returning to the Home position on the steering wheel. A switch was depressed on the steering wheel between reaches with the participant’s right hand to collect the reach time and to ensure a consistent origin for each reach. The order of targets to be reached was randomized to reduce

36

anticipatory effects, and the participants responded to the experimenter’s verbal indication of the next target to be reached. Reaches were performed approximately every ten seconds, allowing several seconds of rest between reaches. Reaches were not temporally constrained, and participants were encouraged to perform the reaches at normal, “comfortable” speeds. A practice session was provided so that the participants could become familiar with the experimental protocol and task requirements. Based on the experience of their practice sessions, participants rated the perceived difficulty of each reach after its successful completion. Requiring difficulty ratings was primarily used to increase the participant’s attention to the reaching task. Ratings were given in terms of “Low”, “Medium”, or “High”, which were coded as 1, 2, or 3 respectively. Frequency sweeps study. Participants were instructed to extend their right arm in a pointing posture and to attempt to maintain its position in either the lateral, forward, or overhead directions. In each direction, two distances were used: Near and Far. Near postures constituted an outstretched fingertip held at approximately shoulder-height with comfortable flexion of the elbow, and without torso flexion. Conversely in the Far postures, participants were instructed to maximally extend their arm. In the forward and lateral directions, the torso was flexed approximately 30° (from vertical) in the direction of the reach. In overhead pointing postures, participants were asked to maximally extend the arm upward without altering the posture below the waist, constraining the postural adjustment to the upper body. The RMS was programmed with a single-pass, 72-second logarithmic frequency sweep, or chirp signal, with the acceleration held constant at 0.5 g (p-

37

p). The sweep began at 0.5 Hz and transitioned through 32 Hz; the frequency doubled every twelve seconds. Under constant peak acceleration, the linear displacement of the RMS cab decreases as the frequency increases. DATA ANALYSIS Motion data were exported from the VICON workstation and imported into the Jack software. A three-dimensional CAD model of the RMS cab was also imported to facilitate visualization of the motion data. Digital human figure models were scaled to match the subjects’ primary body dimensions. The cab-relative movements of the subject’s fingertip were used to analyze the movement time and endpoint accuracy of the reaches. Movement times were measured between the initial movement beginning the reach and the moment the fingertip touched the target. The frame preceding the onset of movement was included as time zero. Reach accuracy was measured in terms of the effective target width (ETW), defined as the target size required to ensure that endpoint locations within approximately two (2) standard deviations of the mean would successfully hit the target (Welford 1968, MacKenzie 1992). This equates to 96% of the normally distributed spatial dispersion of reach endpoints. Correlation coefficients were computed to reveal trends in the data, and the qualitative findings provide the basis for the studies described in the following chapters. Statistical models were represented by mixed linear models in JMP® 12.0 (SPSS, Inc., Chicago). ANOVA was performed, for which results were significant at the p < 0.05 level.

38

Results Hypothesis 3.1: The effective target width of point-to-point reaches under whole-body vertical vibration is significantly greater at frequencies between 4 Hz and 6 Hz. The participants' exhibited large variability in the ETW when performing the reaching tasks at 4, 5, and 6 Hz, under a constant acceleration of 0.8 g p-p. Table 3-1 indicates that the target location and vibration frequency were both significant main effects, and significant two-way interactions as well. Table 3-1. ANOVA results for the fingertip variability at the target during vertical sinusoidal vibrations.

Source

N

DF

S.of S.

F Ratio

Prob > F

Subject

1

1

0.15

1.13

0.296

Target

5

5

25.1

36.8

< 0.0001

Frequency

6

6

107.2

131.2

< 0.0001

Subject * Target

5

5

0.80

1.17

0.348

Subject * Frequency

6

6

0.51

0.63

0.706

Target * Frequency

30

30

10.3

2.53

0.007

The Mean ETW values are presented in Figure 3-7, in which the effects of target locations and vibration frequency can be more clearly seen. The average standard deviation across all conditions was 23% of the mean. Reach accuracy improved above 6 Hz, and the ETWs at 10 Hz were similar to those observed in the stationary condition. This frequency-dependent relationship of mean ETW confirms performance degradation near the principal resonant frequency of the body, and provides evidence that this resonance directly affects the manual control of the hand during reaching tasks.

39

Effective Target Width (cm)

5 4 3 2 1 0 0 Hz

2 Hz

4 Hz 5 Hz 6 Hz Input Frequency

8 Hz

Overhead (T1)

Glove box (T5)

Radio (T4)

Steering Column (T3)

Floor (T6)

Speedometer (T2)

10 Hz

Figure 3-7. Mean fingertip variability at the target (an estimate of the effective target width), as a function of vibration frequency and target location.

Hypothesis 3.2: Target location significantly affects the effective target width of point-to-point reaches under vertical whole-body vibration. Figure 3-7 shows that reaches to some target locations are more susceptible to vertical vibrations than other locations. The largest difference is between the close reach directly forward (Speedometer), which has the smallest ETW (~2 cm), and the Overhead reach which requires a target more than twice as large (~4.8 cm). In fact, there appears to be a proportional relationship between target locations, independent of the vibration frequency. For example, ETWs for the Overhead and Speedometer reaches appear to have close to a 2.5 to 1 ratio for each frequency, inclusive between 2 Hz and 8 Hz.

40

Hypothesis 3.3: Perceived reach difficulty is correlated with the effective target width. A correlation (r = 0.89) was found between reported perceived difficulty scores and the effective target width (Figure 3-8), indicating that the principal resonant frequencies of 4-6 Hz had similar effects on perceived difficulty as they had on ETW. Movement times also correlated with both the reported difficulty ratings (r = 0.74), and the ETW (r = 0.59). Individual difficulty ratings had little variability between subjects, for which only 6 of the 42 responses were different between subjects. Difficulty ratings for all of the reaches in the stationary condition were “Low”, as were those for the reaches performed at 10 Hz. Subjects gave the shoulder-high, forward reach to the “Speedometer”, the lowest mean difficulty at 1.5 across the tested frequency range. Conversely, the “Overhead” and “Glove box” yielded the highest average scores, both with difficulty ratings of 2.3 (0.9). Reaches to the “Radio” were also difficult, particularly in the 4–6 Hz resonance band, with an overall rating of 2.1 (0.9). None of the reaches to the Speedometer were rated above 2 (medium difficulty), most likely due to the short travel distance required to reach the target.

41

Difficulty Rating

Movement time

4

40

3

30

2

20

1

10

0

0 Static

2 Hz

4 Hz

5 Hz

6 Hz

8 Hz

Increase in MT (%)

Endpoint variability (cm) and Difficulty ratings

Fingertip Variability

10 Hz

Frequency Figure 3-8. Illustration of correlation of the frequency effects related to fingertip variability, subjective difficulty ratings, and relative increases in movement times.

Hypothesis 3.4: Fingertip variability of maintained pointing postures during the vertical frequency sweep is greater when the arm is fully extended than in postures where the extended hand is closer to the body. Figure 3-9 shows the variability of the extended fingertip pointing postures, as known as the effective target width, as a function of vibration frequency. Solid lines correspond to the 180-point (3 sec) moving average of the fingertip movement, and the dashed lines show the standard deviation of the variability. Horizontally extended arm postures also showed significantly higher variability at 2 Hz than when the arm was extended overhead. Though the platform vibrations were strictly in the vertical axis, horizontal torso movements were stimulated by the 1-2 Hz vibrations; frequencies at which the torso is known to resonate in the horizontal direction (Whitman and Griffin 1978).

42

60 40

Effective Target Width (mm)

20 0

Forward - Near

Forward - Far

Lateral - Near

Lateral - Far

60 40 20 0 60 40 20 0 0.5

1

2

4

8

16

32

0.5

1

Overhead - Near

2

4

8

16

32

Overhead - Far Frequency (Hz)

Figure 3-9. Variability of effective target width under vertical frequency sweep.

The inclusion of the torso in a maintained pointing posture, and thus a longer moment arm of the extended hand with respect to the fulcrum at the waist, may cause the increased movement of the fingertip in response to the vibration stimulus. However, Figure 3-9 shows that while there was noticeable variability of the effective target width during the frequency sweep, there was not a consistent effect with respect to distance. While all of the pointing postures exhibited increased variability around 4 Hz, the horizontally extended arm postures exhibited a larger range of increased variability between 1 and 4 Hz. The sensitivity of the seated human to horizontal vibrations in the 1-2 Hz range (Whitman and Griffin 1978) seems to be excited even under vertical motion. The

43

vertical motion may induce pitching motion of the pelvis as Sandover (1983) suggests, which would explain the observed increase in the effective target width at 2 Hz. Overall, the general similarity of the frequency response, regardless of direction and distance, support a proposition of McRuer et al. (1965), which suggests that people are capable of adjusting their joint stiffness parameters to modulate transmissibility in order to obtain a nearly constant output of the vibration feedthrough of the end effector in a pointing task.

Discussion This study confirmed that whole-body vertical vibration will elicit significant biodynamic effects, and that three-dimensional reaching movements could be successfully recorded by an optical motion capture system within a dynamic ride motion environment. The sampling frequency of 60 Hz proved to be sufficient to record accurate marker data in the dynamic environment. Visual inspection of the digital reconstruction of marker locations revealed that postures and movements appeared natural and smooth, despite the volatility of the ride motion environment. Figure 3-7 shows that overhead reaches, or extended horizontal reaches (i.e. Glovebox) are highly susceptible to vertical vibrations using this metric. This preliminary result suggests that pushbutton targets will need to be much larger to maintain stationary postures if certain frequencies (i.e. 4-6 Hz) are expected. More specifically, while reaches performed in the stationary cab were highly accurate (ETW ~ 0.5 cm), reaches performed under vibrations within the range of 4-6 Hz exhibited much higher fingertip variability (ETW ~ 2.5 cm). Further the

44

overhead reach would require a pushbutton target that is nearly 5 cm (~ 2 inches) to ensure an accurate reach under such vibration conditions. One new finding that relates to this upward pointing posture was observed when subjects attempted to hold their arm overhead during the frequency sweep. During the 4-6 Hz resonance frequencies of the torso under vertical motion, the participant’s lumbar spine and the torso flexes, inducing a fore-aft motion of the finger during the overhead reach. Each subject reported increased difficulty ratings when keeping the arm stationary during these sessions. Accordingly, these upward reaches exhibited nearly twice the variability between 4 and 6 Hz than at other frequencies and received “High” difficulty ratings. Furthermore, the spatial variability of the extended arm and fingertip show that manual performance of the uncoupled hand is similarly affected by these vibration frequencies. A high correlation was found between reported difficulty ratings and the effective target width (R > 0.89), showing that the principal resonant frequencies of 4-6 Hz have similar effects on perceived difficulty as they have on reach accuracy. Movement times also were correlated with both the reported difficulty ratings (R > 0.74), and the ETW (R > 0.59). The correlations between perceived task difficulty, reach accuracy, and movement times suggest that there is a consistent biodynamic effect of vibration frequency on the manual performance of seated reaching tasks. This biodynamic effect of vertical vibration depends on the vibration frequency and reach direction, and will be further evaluated with respect to lateral and longitudinal sinusoids (Chapter 4), and six random degree-offreedom ride motion (Chapters 5 and 6).

45

All of the pointing postures held during the frequency sweeps exhibited significant peaks at around 4 Hz, although there was no observable difference between the variability associated with reach distance. Both Near (~30 cm) and Far (~60 cm) postures resulted in ETWs between 30 and 50 mm. Military specifications (MIL-STD-1472) require that in-vehicle pushbuttons are sized between 9.5 and 50 mm (quite a large range!), which does accommodate these findings, but the standard does not provide necessary assistance on how task parameters affect the preferred target size within that range. Woodson et al. (1992) provide some additional assistance on target characteristics based on the location within the vehicle; however their recommendations do not appear to incorporate the effects of ride motion perturbations on reaching movements presented in this and following chapters. Chapters 4 and 5 present results pertaining to the preferred location, shape, and orientation of pushbutton targets as a function of the direction, frequency, and magnitude of vehicle ride motion. The principal resonance of the seated person has been shown to be between 4 and 8 Hz, and both the reaches to in-vehicle targets and the postures held during frequency sweeps exhibited increased difficulty and large ETWs within this frequency range. However, there appears to be a difference in the ETW, depending on whether the subject is reaching out to pushbutton target, or simply maintaining the terminal posture under vibration. There is a striking difference in magnitudes of ETWs between reaching and maintaining postures under the same vibration frequencies. To illustrate a similarity, Figure 3-7 indicates that an overhead reach at 4 Hz vibration will result in an ETW of nearly

46

5 cm (~2 in). Likewise, maintaining the same terminal posture of the overhead reach resulted in an ETW also estimated above 4 cm. Conversely, while the location of the fingertip during the Forward Near condition (Figure 3-9a) is very near the location of the “Speedometer” target, the ETW of the maintained posture (~4.5 cm) was quite different than the terminal reaching posture (~2 cm). This discrepancy in ETWs may result from differences in neuromuscular control during these tasks. In the impulse-timing theory (Woodworth 1899), muscle activity is constantly changing during target-directed movements, which may make the system more responsive to needed modifications. Reaching movements rely on this timing of agonist and antagonist muscle activation to manipulate the arm and fingertip from a neutral posture to the target location. Conversely, in the equilibrium-point hypothesis (Feldman 1966, 1986), the spring and dashpot parameters of the agonist and antagonist muscle systems are fixed and utilize muscle coactivation that adjusts the arm back to the desired equilibrium position. The feedback control processes that are utilized in these scenarios are further evaluated in Chapter 6. These preliminary findings suggest that design guidelines for pushbutton controls in land-vehicles should consider both ride-motion characteristics and reach direction and distance (i.e. target location). In a ride-motion environment, the performance of reaches to a button of a particular size, in terms of speed or accuracy, is likely to depend on the button’s location in a manner that is different from that observed in static trials. This suggests that design guidelines based on static data should be re-examined for application to moving vehicles.

47

Chapter 4 Study 2: Analysis of the effects of vibration frequency, magnitude, and direction on movement times Certain types of vehicle motions can perturb hand movement trajectories. The objective of the work reported in this section is to evaluate the main and interactive effects of vibration frequency, magnitude and direction on movement times associated with in-vehicle reaching tasks.

Background Vehicle operators are required to perform a variety of reaching tasks while the vehicle is in motion. As was shown in chapter 3, vertical sinusoidal vibrations disrupt the operator’s ability to reach out and press pushbutton controls. In general it is expected that the difficulty of controlling the movement of the arm in a dynamic environment becomes increasingly challenging as the arm is extended away from the body, as described earlier in chapter 3, thus resulting in higher force exertions to counteract the perturbations and the decreased stability. It has been shown that the distance of the reach and the shape/orientation of controls and displays to which a person reaches affects movement times (MacKenzie & Buxton 1992), which is a function of the difficulty of fast reaches (Fitts 1954). Reed et al. (2003) also showed that target location has a significant effect on the perception of reach difficulty. Results presented in chapter 3 show a correlation exists between reach difficulty and movement time.

48

Soechting & Lacquaniti (1981) have shown that the trajectories (and other characteristics) of pointing tasks are highly repeatable, and are independent of the hand velocity. When rapid reaches are performed, reach trajectories tend toward a straight line (Morasso 1981), but as temporal constraints are relaxed, trajectories exhibit more curvature, and have increased variability across subjects (Osu et al. 1997). In common tasks, precise trajectories are not required, as only the endpoint of a reach is constrained. The time to complete a reach (movement time) and the endpoint accuracy are movement characteristics that can be compared across subjects and trials. Significant perturbations resulting from ride motions would be expected to create large deviations from the intended trajectory, and may correlate to measures of reach difficulty. Results presented earlier in chapter 3 indicated that degradation in manual performance under vertical vibration was most noticeable in vertical movements of the hand under vertical vibrations between 4 and 6 Hz. This frequency range is known to induce a principal resonance of the upper body in the seated posture (Coermann 1962). It is also known that vertical and horizontal vibrations affect the seated operator differently. The principal resonance of seated vehicle operators under horizontal vibrations has been estimated to be between 1 and 2 Hz (Whitman and Griffin 1978). The effect of vibration frequency is closely coupled with its magnitude, where vibrations within a resonance band may not affect performance if the magnitude is not large enough, or vice versa. In this context, the specific interaction between frequency

49

and magnitude is known to be nonlinear (Hinz & Seidel 1987; Mansfield & Griffin 2000). Based on the preceding, a systematic analysis of the influence of direction, frequency, and magnitude of vibration transmission through the torso should provide information to discern preferrable locations of controls and displays within the vehicle cockpit. The present study investigates the movement times and endpoint accuracy of in-vehicle reaches when single-axis sinusoidal motion in the lateral, fore-aft, and vertical directions are applied to a seated person. Using motion capture of the end effector (right index fingertip) trajectory, an understanding of how the CNS develops, executes, and alters motor programs under dynamic ride motion is investigated. The analysis presented here focuses primarily on the duration of the movements and the variability in the trajectories during ride motion disturbances. The experiment was designed to test the following hypotheses: Hypothesis 4.1: Vibration magnitude will significantly affect movement times to distance-normalized targets, Hypothesis 4.2: The frequency and direction of vibration will have a significant interaction with respect to movement times, Hypothesis 4.3: The direction of vibration and reach direction will have a significant interaction with respect to movement times, and Hypothesis 4.4: The deviation of three-dimensional reach trajectories away from a straight line will be proportional to the reach distance.

50

Materials and Methods The Ride Motion Simulator (RMS) Laboratory at the U.S. Army Research, Development, and Engineering Command (RDECOM) was used to simulate single-axis sinusoidal displacements in three primary axes: vertical, lateral, and antero-posterior. Based on the literature and results presented in chapter 3, vibration frequencies were chosen in discrete 2 Hz intervals from 2 Hz to 8 Hz, surrounding the resonant frequency band of vertical vibration (4-6 Hz). A stationary condition was also used as a baseline for the motion conditions. Vibration

magnitudes

are

given

as

root

mean

square

(r.m.s.)

accelerations, and were based on comfort curves of a seated human’s sensitivity to motion direction (ISO 2631-1), and were nominally identified as “low”, “medium”, and “high”. The lateral and fore-aft accelerations were 0.693 m/s2 (low), 1.387 m/s2 (medium), and 2.08 m/s2 (high). Likewise, the vertical accelerations were 1.387 m/s2, 2.427 m/s2, and 3.467 m/s2 respectively. Applying the frequency weightings from the ISO 2631 standard yields weighted accelerations ranging from 0.173 m/s2 to 3.467 m/s2. These values comprise a range of accelerations from smooth rail transportation, 0.8 m/s2 r.m.s. (Johanning et al. 1991), to rough military tanks, exceeding 3.0 m/s2 r.m.s. (Maksimovic 1987). The “rough” accelerations fall below the fatigue-decreased proficiency boundary (ISO 2631-1). In each condition, subjects performed reaching tasks to eight pushbutton targets located in the right-hand reach envelope. Twelve subjects (6 men, 6 women) participated in this study and were restrained with a lap belt only to allow

51

unrestrained movement of the torso. Summary metrics of general subject characteristics are provided in Table 4-2. Table 4-2. Participant summary: Stature (cm), Weight (kg), Age (yr). Male (n=6)

Female (n=6)

Stature

Weight

Age

BMI

Stature

Weight

Age

BMI

Mean

176.5

75.8

28.5

24.9

169.3

71.2

28.2

22.9

SD

5.5

11.7

8.4

3.0

6.4

4.4

5.4

2.6

Range

170-185

66-98

22-45

68-78

22-36

18.5-26.5

21.6-30 160-180

EXPERIMENTAL SETUP Subjects were seated in the RMS cab with hands placed initially at the home position on the steering wheel. Reflective markers were placed on anatomical landmarks as described below. A “Home” switch was located at the 2 o’clock position on the steering wheel, which was depressed by the subject before movement onset and after completion of each reach. Analog signals were recorded for the Home switch and the eight pushbutton targets to calculate reaching times. Four of the eight targets were reachable from a normal seated posture, without having to bend the torso (Figure 4-1). These targets were located in each of the three primary axes, and one additional reach at a 45° diagonal between the lateral reaches. Pushbutton targets were 1.27 cm (0.5”) square, and located within the right-hand reach envelope. These targets are referred to by their number and qualitative description: 1. Overhead (OH), 3. Forward-Near (FN), 5. Forward-Lateral (FL), and 7. Lateral-Near (LN).

52

1

3 5

7

Figure 4-1. Depiction of reaches to targets that did not require use of the torso.

Nominal postures to the remaining four targets are depicted in Figure 4-2. All reaches to these targets required a moderate amount of torso assistance (i.e. flexion or lateral bending), which was calculated from the motion capture system data. These reaches had extended hand motions forward and lateral, and 45° upward. The targets for these are referred to by their number and qualitative description: 2. Forward-Up (FU), 4. Forward-Far (FF), 6. Lateral-Up (LU), and 8. Lateral-Far (LF).

53

2

4

6

8

Figure 4-2. Depiction of reaches to targets that required torso movements.

Figure 4-3 shows the configuration of the RMS cab with a subject seated in the home posture. Targets were rigidly attached to the aluminum alloy cab frame, and were not sensitive to the vibration stimulus.

54

Figure 4-3. Participant seated in the RMS cab with hands in the home position on the steering wheel.

A twelve-camera VICON 524 motion capture system, sampling at 60 Hz was used to record the movement of twenty-nine (29) reflective markers placed on key anatomical landmarks of the participant (Figure 4-4). The cameras were positioned around and above the RMS cab, resolving much of the potential occlusion. The VICON system also recorded analog data from the eight pushbutton targets and the home switch located on the steering wheel.

55

Figure 4-4. Digital human with reflective markers in the home resting posture.

EXPERIMENTAL TASKS Right-handed reaches were executed from the steering wheel to each of the eight pushbuttons targets that were activated with the right index fingertip. A five-minute training period was provided to help the subject become comfortable with the task requirements. Movement times were measured during “self-paced” and “rapid” reaches in the stationary and vibration conditions under the single-axis sinusoidal vibrations described earlier. In the stationary condition, subjects performed five “self-paced” and five “rapid” reaches to each target. “Self-paced” reaches were performed at a relaxed speed, where the subjects were instructed to “reach out and press the center of the target with the right index fingertip,” which provided a measure of intersubject differences in the nominal time taken to accurately perform the reach. Conversely 56

for “rapid” reaches subjects were instructed to “press the button as quickly as possible,” which provided a measure of the minimum time during which the participant could successfully complete the reaching task in a stationary environment. After completing each reach, subjects were required to return their right hand to the steering wheel, depress the Home switch, and wait for instructions on the next reach. Reaches were executed in response to voice commands, and each trial was concluded when the target had been successfully pressed five times. In the vibration conditions, subjects performed only “self-paced” reaches. The experimental block (direction x frequency x magnitude) was randomized, and subjects performed reaches to each of the eight targets without replication. Reaches were performed under three vibration magnitudes (low, moderate, and rough) as described earlier. DATA ANALYSIS As expected, a preliminary analysis revealed a high intersubject variability in movement times for the “self-paced” reaching tasks, as this condition only emphasized accuracy. Consequently, motion time data were transformed into a log-normal distribution (Faraway 2004), which has been shown to have a good fit for similar human movement studies (Griffin & Whitman 1978). Mean movement times for all targets is presented in Figure 4-5, while Figure 4-6 shows the transformed data and the normally distributed values that result.

57

Movement time (ms)

1200 1000 0 Hz

800

2 Hz

600

4 Hz

400

6 Hz 8 Hz

200 0 1

2

3

4

5

6

7

8

9

10 11 12

Subject Figure 4-5. Variability in movement times between participants for self-paced reaches in which subjects were only instructed to successfully push the target.

lognormal MT

0.2 0.1 0 -0.1 -0.2 1

2

3

4

5

6 7 Subject

8

9

10

11

12

Figure 4-6. Lognormal distribution of movement times showing constant variance.

In this study, movement times were used as the primary performance metric. If the intended target was not activated during a reach, the reach was deemed unsuccessful (i.e. the subject missed the target), and the trial was repeated under the same conditions, randomly reinserted in the experimental block. A measure of the trajectory variability was assessed with respect to the distance of the fingertip perpendicular to the direction of fingertip movement, also

58

known as off-axis or “cross-track” error. This minimum distance between the mean trajectory of reaches performed in the stationary environment and the trajectory of a reach performed under vibration is known as the Frechet distance. Statistical models were represented by mixed linear models in JMP® 12.0 (SPSS, Inc., Chicago) and R-2.0.1 for Windows (Ihaka and Gentlemen, 1996). Analysis of variance with repeated measures (ANOVAR) was performed, for which results at the p < 0.05 level are referred to in the text as “significant”, and effects at the p < 0.01 level are referred to as “highly significant”. Forward stepwise regressions were performed for the factors listed in the ANOVAR tables, beginning with the most significant factor, and p < 0.05 was used as the inclusion criterion. Linear regression models were generated based on the least squares estimates of the significant factors identified by the regression method. Significance of the factor levels and interactions were tested using a TukeyKramer analysis of unbalanced pairwise comparisons at the p < 0.5 level.

Results Table 4-3 shows the main effects and interactions that significantly affect the movement times (MT) of temporally-unconstrained reaches. As expected, the target location significantly affected MT (p < 0.01), due to the effect of reach distance (Crossman and Goodeve 1983); however normalized movement times also were significantly affected by the azimuth and elevation of the target location, with respect to the initial orientation of the body, as described below. Across vibration conditions, normalized movement times for reaches that required use of the tors, were 11.6% longer than reaches that did not require

59

torso movement. Vibration frequency and magnitude also were significant main effects. A least squared fit of the significant factors identified through forward stepwise regression accounted for 57% of the variation in MT. Table 4-3. ANOVA results for lognormal movement times for distance-normalized reaches without temporal constraints.

Source

N

DF

S.of S.

F Ratio

Prob > F

Target (T)

7

7

12.1

433

0.0000

Vibration Direction (VD)

2

2

0.07

9.07

0.0001

Vibration Frequency (VF)

3

3

0.17

14.2

< 0.0001

Vibration Magnitude (VM)

2

2

0.12

15.1

< 0.0001

T * VD

14

14

0.10

1.80

0.034

VD * VF

6

6

0.10

4.06

0.0005

VF * VM

6

6

0.12

5.03

< 0.0001

Hypothesis

4.1:

Vibration

magnitude

will

significantly

affect

movement times to distance-normalized targets.

Table 4-3 shows that vibration magnitude significantly affects the MT (mean ± standard error) of temporally-unconstrained reaches. Across all conditions, mean MTs were significantly higher in the Moderate (638 ± 7.0) and Rough conditions (640 ± 6.9) than in the Mild condition (623 ± 6.8), though there was not a significant difference in MT between the Moderate the Rough conditions. The interaction of vibration frequency and magnitude also was significant. Figure 4-7 shows the mean MTs for each combination of vibration frequency and vibration magnitude, where the mean MT of reaches performed in the stationary condition is represented by a dashed line. Tukey-Kramer (T-K) analyses of pairwise comparisons were performed on mean MTs for reaches performed in the stationary condition, and reaches performed under each combination of vibration frequency and vibration magnitude. The T-K analyses did not show a significant 60

difference in mean MT between reaches performed under Moderate vibration at 6 Hz and the stationary condition; however MTs were significantly longer for all combinations of vibration frequency and magnitude than for the stationary condition. Under Mild vibration, MTs were significantly longer for reaches performed in the 2 Hz condition (663 ± 8.5 ms) than for the other frequencies, which averaged near 640 ms. Mean MTs for reaches performed under Moderate vibration were not statistically different between 2 Hz and 8 Hz (~664 ms), though mean MTs were significantly longer at 4 Hz (685 ± 9.4 ms), and significantly shorter at 6 Hz (630 ± 8.9 ms). Finally, the Rough condition did not induce a significant difference in mean MT between 6 Hz and 8 Hz, but reaches were significantly longer at both 2 Hz (680 ± 8.7 ms), and at 4 Hz (700 ± 9.4 ms). The interactions between target location and vibration frequency, and target location and vibration magnitude were not significant. 725

Movement times (ms)

Mild

Moderate

Rough

700 675 650 625 600 2 Hz

4 Hz

6 Hz

8 Hz

Frequency Figure 4-7. The interaction effect of vibration frequency and vibration magnitude on mean movement times. The dashed line represents the mean movement time for all reaches performed in a stationary environment.

61

Hypothesis 4.2: The frequency and direction of vibration will have a significant interaction with respect to movement times.

The ANOVA results presented in Table 4-3 above shows that the vibration direction, vibration frequency, and their interaction were all significant. Figure 4-8 illustrates the effects of vibration direction and frequency on mean movement times across all targets and vibration magnitudes. T-K comparisons showed that the interactions were not significant difference between reaches performed in the stationary condition and those performed under lateral vibrations at 8 Hz. However all other combinations of vibration frequency and vibration direction induced significantly larger mean MT than in the stationary condition. Mean MTs for reaches performed under Lateral vibration were not significantly different between 4 Hz and 6 Hz (~649 ms); however reaches were significantly longer at 2 Hz (687 ± 8.7 ms), and significantly shorter at 8 Hz (627 ± 8.7 ms) than reaches performed under middle frequencies of vibration. The Fore-aft vibration condition resulted in mean MTs that were divided into two groups: reaches at 2 Hz (679 ± 9.3 ms) and 4 Hz (681 ± 9.1 ms) were significantly longer than reaches at 6 Hz (653 ± 8.6 ms) and 8 Hz (656 ± 8.7 ms). Conversely, mean MTs were not statistically different between vibration conditions in either group. Under Vertical vibration, mean MTs were not significantly different between reaches performed at 2 Hz (673 ± 8.7 ms) and at 6 Hz (668 ± 9.3 ms). Reaches performed under 4 Hz vertical vibration resulted in significantly longer mean MTs (695 ± 9.4 ms), while reaches under 8 Hz vibrations were significantly shorter than in the other vibration frequency conditions (651 ± 8.5 ms).

62

725

Movement times (ms)

Lateral

Fore-Aft

Vertical

700

675

650

625

600 2 Hz

4 Hz

6 Hz

8 Hz

Frequency Figure 4-8. Effect of vibration frequency and vibration direction on mean movement times.

Hypothesis 4.3: The direction of vibration and reach direction will have a significant interaction with respect to movement times.

As illustrated in Figure 4- and Figure 4-2, five target locations required a general reach direction that coincided with the direction of a vibration condition. For example, the Overhead target was located directly above the shoulder; hence the reach direction to this target coincided with the Vertical vibration direction. Similarly, the reach direction to the Forward Near and Forward Far targets coincided with the Fore-aft vibration, while reaches to the Lateral Near and Lateral Far targets were performed along the same axis of the Lateral motion condition. Figure 4-9 shows the interactions between target location and vibration direction on the normalized movement times for all targets, where vibration contributed to increased movement times in all conditions compared to reaches to these targets in the stationary condition. The targets are ordered with respect to reach distance, from the shortest travel distance (left) to the longest (right).

63

T-K pairwise comparisons were used to evaluate the interactions between target location and vibration direction for these five targets. There was not a significant difference in the mean MTs for reaches to the Overhead or Forward Far targets across the vibration directions. There was however a significant decrease in mean MT when reaching to the Lateral Near (671 ± 12.3 ms) target under lateral motion, than reaching to the same targets under vibrations perpendicular to the reach direction (i.e. fore-aft, vertical), 709 ± 12.4 ms and 703 ± 12.8 ms, respectively. Likewise, there was a significant decrease in mean MT for reaches to the Lateral Far (782 ± 13.3 ms), compared to the Fore-aft (829 ± 13.7 ms) and Vertical (813 ± 14.7 ms) conditions. A similar effect was observed for reaches to the Forward Near target, where mean MTs were 476 ± 12 ms under Fore-aft vibration, as compared to the significantly longer mean MTs under Lateral (495 ± 12.4 ms) and Vertical (489 ± 12.6 ms) vibration. In summary, across conditions reaches that coincided with the vibration direction were on average 5% longer than in the stationary condition, while reaches in a perpendicular direction to vibration motion were on average 8% longer than in the stationary condition. A paired t-test indicates that this difference between the means is significant (p < 0.05), though the 3% difference in MT for these movements only equates to approximately 20 ms.

64

120%

Lateral

Fore-aft

Vertical

Normalized MT

115%

110%

105%

100%

95% Forw ard Near

Forw ard Far

Forw ardLateral

Forw ardUp

Lateral Near

Overhead Lateral-Up Lateral Far

Target location Figure 4-9. Interaction vibration direction and target location on movement times, normalized by movement times in the stationary condition.

Hypothesis 4.4: The deviation of the reach trajectory away from a straight line is proportional to the reach distance.

Curvature of a reach trajectory was estimated by the Frechet distance. Figure 4-10 presents the ratios between the Frechet distance and the reach distance (F/D ratios). T-K comparisons revealed that the F/D ratio for the Forward Near target (17.0 ± 0.3%) was significantly greater than all of the other targets. Similarly, reaches to the Forward Far target (9.8 ± 0.4%) resulted in F/D ratios that were significantly greater than the remaining six targets. The only additional significant difference in F/D ratios was between the Lateral-Up target (10.3 ± 0.4%) and the Lateral Far target (14.0 ± 0.4%), across all vibration conditions. One possible cause of significantly greater ratios for the two forward reaches was that the presence of the steering wheel may have induced increased curvature of the fingertip trajectory in order to avoid the hand hitting

65

the steering wheel. The unobstructed reaches to the other six targets exhibited mean Frechet distances that averaged approximately 7.3% (± 0.4%) of the reach distance. Proportion of Frechet deviation to reach distance

Frechet / Distance

0.20

0.15

0.10

0.05

0.00 Forward Near

Forward Far

FwdLateral

ForwardUp

Lateral Near

Overhead

LateralUp

Lateral Far

Target Location (L -> R by distance) Figure 4-10. Deviations to the reach trajectory from a straight line are very similar, with exceptions of the Forward Near and Forward Far targets, to which reaches were required to avoid the steering wheel.

A backward stepwise regression (removal criterion: p > 0.1) was performed, using the same factors given in Table 4-3, to analyze the significance of their respective main and interactive effects in predicting F/D ratios. Table 4-4 shows the resulting least squares model, where only the target location and vibration frequency were significant main effects. None of the interactions were significant. With regards to the vibration frequency, only the 2 Hz condition (9.4 ± 0.2%) resulted in a significant difference in F/D ratios, when compared to the 4 Hz, 6 Hz, and 8 Hz conditions – 8.8%, 8.6%, and 8.4% respectively.

66

Table 4-4. ANOVA results for ratio of Frechet distance to reach distance.

Source

N

DF

S.of S.

F Ratio

Prob > F

Target (T)

7

7

2.66

104

< 0.0001

Vibration Frequency (VF)

3

3

0.04

3.33

0.019

Discussion Whole-body vibration (WBV) results in increased movement times of temporally-unconstrained reaching tasks due to the direction, frequency, and magnitude of single-axis sinusoid vibrations. Specifically, lognormal movement times (MT) increased by 6.0% for low magnitudes, 10.0% for moderate, and 16.5% for rough magnitudes of vibration when each is compared to MT for reaches performed in a stationary condition across all conditions. However, the interactions between the direction, frequency, and amplitude of WBV on MT are complex, and thus predicting MT based on these factors remains a challenge. Although Fitts law (Fitts 1954) is not valid for temporally unconstrained movements, the significance of the vibration main effects and interactions in the above analyses suggest that an additional component may be useful in Fitts’ law to incorporate possible effects of vibration on movement times. This aspect will be further discussed with respect to temporally constrained reaches in chapter 5. An additional consideration for predicting MT under vibration is the discontinuity observed in reach distance. Recall that four targets (2, 4, 6, and 8) required recruitment of the torso to complete the reach, where these reaches required 11.6% more time to complete than reaches that did not require use of the torso. Despite the large variability in nominal movement speed between subjects, vibration induced significant increases in the movement time required to

67

complete these temporally-unconstrained reaching tasks. Vibration resulted in increased movement times by as much as 25%, however the effect was dependent on the direction, frequency, and magnitude of the vibrations. Consistent with published reports, resonant effects were observed about 4 Hz under vertical vibrations (Coermann 1962), and 2 Hz for horizontal vibrations (Whitman and Griffin 1978). However these known biodynamic effects for stationary seated postures have been extended in this chapter to the prediction of movement times for temporally-unconstrained reaches under vibration. Vibration feedthrough displaces the fingertip away from the intended goal, which requires additional time to make compensatory movements to successfully press the target. Increases in vibration magnitude at these resonant frequencies further excite the biomechanical linkage system, although some results have shown a dampening effect at higher magnitudes (Fairley & Griffin 1989). Vehicle designers should ensure that transmitted vibrations have minimal power near these resonant frequencies, and utilize a systems-level approach to vehicle design that evaluates the vibration feedthrough through the tires, wheels, suspension, and seating. There was also a significant difference in movement times when the reach direction coincided with the vibration direction. Lateral reaches exhibited the shortest movement times under lateral vibrations. Reaches to the Forward Near target also were shorter under fore-aft vibrations, however extended forward reaches may have been more difficult due to the pitching motion of the pelvis that has been associated with vertical and fore-aft motion (Wilder et al. 1982; Pope et

68

al. 1991). Similarly, there was not a significant difference observed in the movement times for overhead reaches. The posture seems particularly susceptible to vibration feedthrough, as vibrations in each perpendicular axis perturb the arm, and vertical motions induce the same pitching of the pelvis observed under fore-aft vibrations. Compared to reaches in the stationary condition, movement times were 6.6% longer for reaches performed in a direction coinciding with the vibration, and 11.1% longer when vibration was perpendicular to the reach direction. A proportional relationship was shown between the Frechet distance and reach distance for unobstructed reach trajectories, where the deviation of fingertip from a straight line was approximately 8% of the reach distance. However reaches to the forward targets exhibited increased Frechet values due to avoiding the steering wheel during the reach. Meyer et al. (1988) provide some support that trajectory corrections can produce curvature in order to complete spatially-constrained movements. This exocentric strategy assumes that this fine motor adjustment is satisfied through the monitoring and adjustment of the endeffector (i.e. fingertip) as it approaches the target (Elliott et al. 1999, Sarlegna et al. 2004). On the other hand, Soechting & Flanders (1989a) suggest that reaching movements may be planned and initially executed in an egocentric, or joint-based, reference frame. This would place less emphasis on the fingertip trajectory, which could result in early deviations to the intended reach. These deviations and the subsequent feedback-based corrections are investigated and discussed in chapter 6.

69

Chapter 5 Study 3: Effects of ride motion on movement time and accuracy This study explores the interactive effects of terrain-induced ride motion and in-vehicle control design on the successful execution of rapid in-vehicle reaching tasks.

Background The studies described in chapters 3 and 4 generally concern the performance of non-temporally constrained movements. Conversely, a large body of knowledge exists on the expected performance of rapid movements. Fitts’ Law (Fitts 1954) is one of the most well-known movement theories, described in Chapter 2. Although the originally proposed behavioral basis for the Fitts’ model is a source of contention (see Meyer et al. 1988), it remains an excellent empirical predictor of movement times for rapid repetitive tapping and target-aimed movements. The model has also been successfully extended to a variety of movements (Langolf et al. 1976), and even with neurologically-impaired participants (Flowers 1976; Wade et al. 1978). Fitts’ law is one formulation detailing the inverse relationship between movement speed and endpoint accuracy, though many other researchers have investigated this trade-off, most prominently beginning with Woodworth’s dissertation on manual aiming tasks (1899).

70

Crossman and Goodeve (1983) explored the sensory feedback processes involved in reaching tasks and further investigations (Van Beers and Sittig 1996) agreed that vision and proprioception provide feedback information necessary to successfully perform visually-guided pointing movements. The use of this feedback information in continuous manual tracking tasks, specifically with respect to processing time and neuromuscular lags, has been well modeled by McRuer’s crossover model (McRuer and Krendel, 1959; McRuer et al. 1965). In the present context of pointing to a target in a dynamic environment, this neuromuscular control strategy allows the vehicle operator to adjust to the perceived disturbances in order to predict the system state at some future point, an ability referred to as “quickening” (Jagacinski and Flach 2003). Compounding the dynamic problem of movement execution are errors a person makes while planning a motion, which are due in part to an inexact estimation in either the initial egocentric orientation of the body (Soechting and Flanders 1989) or the distance and direction of the target (Messier and Kalaska 1997; Servos 2000; Kim 2005). Additionally, physiological noise in the neuromuscular system is believed to induce errors in executed movements, (Schmidt et al. 1978, 1979; Meyer et al. 1982). Stimulation of the neuromuscular system by whole-body vibration (WBV) is one condition that increases this noise component to further distort the resulting movement (Martin 1981; Gauthier et al. 1981; Gribble et al. 2003). Lastly, movement planning errors may also arise from conflicting vestibular and somatosensory inputs from non-stationary environments, where cross-modal

71

afferent information may alter the CNS’ ability to quickly resolve its initial reference frame, or to compose a mental map of a motion within a rapidly changing environment (Gauthier et al. 1984; Staines et al. 2001; Contreras-Vidal et al. 2004). Spatial disorientation has contributed to as much as 80% of nonmechanical aviation crashes (Li et al. 2001). Similar sensory mismatches with one’s mental map of a planned motion exist in enclosed ground-based vehicles, especially in poor visibility driving situations, where the future state of the environment cannot be adequately predicted. These degradations are thought to be associated with the inexact estimation of orientation and altered movement control (Kagerer et al. 1997), and although humans develop significant manual dexterity at an early age, few gain substantial experience resolving these sensory mismatches. Furthermore, decrements of manual performance, such as lengthened reaction and movement times, can result from the additional cognitive processing that results from the increased complexity of dynamic environments (e.g. maintaining balance and resolving sensory mismatches). This effect is similarly seen when having to reach to a signal target in the presence of distracter targets, where the person must ensure that the generated movement plan will end at the correct destination (Hyman 1953). It has also been shown that an imbalance in vestibular equilibrium affects the CNS’ frame of reference for planning and executing reaching movements (Bresciani et al. 2002; Mars et al. 2003). The observed performance degradation is further supported by studies demonstrating that a moving environment can influence specific aspects of task performance. Examples include, 1) alterations of

72

sensory inputs (Goodwin 1976; Roll et al. 1982, Ghez et al. 1990; Ghez 1991), 2) degradation of movement control by physical perturbations and visual distortions (Polit and Bizzi 1978; Martin et al. 1991; Contreras-Vidal and Kerick 2004), and 3) increase demands on central nervous system processing resuting from postural instability (Redfern et al. 2002). Understanding the particular conditions and mechanisms by which ride motion environments adversely affect operator performance depends on the specifics of both the dynamic environment and the task being performed (von Gierke and Brammer 1996). Despite the inherent variability in a person’s movement plan and execution, it is proposed that the means of the trajectory paths and reach endpoints of repeated, rapid, target-directed reaches in a stationary environment closely approximate the originally intended movement (Soechting and Lacquaniti 1981; Faraway 2000). These mean trajectories can then be used as a reference to compare trajectory characteristics, such as movement time and deviations perpendicular to reach direction, to the same target while being subjected to vehicle ride motion. The specific goal then becomes investigating how land-vehicle ride motions affect the speed and accuracy of rapid, three-dimensional reaching tasks. The decreased ability or inability to predict environmental conditions is predicted to require additional time for movement planning and degraded manual control. That is, the motion environment is expected to be associated with degraded performance, specifically longer reaction times, longer movement times, and decreases in endpoint accuracy of the reaching tasks. The following

73

hypotheses were tested to determine the conditions that seem to most significantly degrade reach performance during ride motion: Hypothesis 5.1: Ride motion will result in longer reaction times, Hypothesis 5.2: Reaction times will be significantly affected by the target location for three-dimensional reaching tasks during vibration, Hypothesis 5.3: Ride motion will result in longer movement times during vibration than in the stationary condition, Hypothesis 5.4: Movement times to digitally presented targets will be shorter than the movement times to physical pushbuttons for threedimensional reaching tasks, Hypothesis 5.5: Variability of reach endpoints will be larger under ride motion than stationary condition, and Hypothesis 5.6: Reaches will be more accurate to physical pushbuttons than reaches to digitally presented targets.

Methods PARTICIPANTS Twenty volunteers (10 men, 10 women) participated in this experiment, free from any knowledge of recent cold or flu symptoms, and had not taken any anti-motion sickness medication. Six men and four women had previous experience riding on the Ride Motion Simulator (RMS), for 7.3 (± 5.1) hours and 4.3 (± 1.7) hours respectively. All subjects were volunteers and signed a consent form approved by both University and US Army Institutional Review Boards. Additional subject characteristics are provided in Table 5-1.

74

Table 5-1. Participant summary: Stature (cm), Weight (kg), Age (yr). Male (n = 10) Mean SD Range

Female (n = 10)

Stature

Weight

Age

Stature

Weight

Age

178.4

79.0

29.9

170.2

66.6

26.2

3.8

9.6

8.4

8.4

11.4

4.6

173-185

68-98

21-45

158-183

51-89

21-37

The present study used the Ride Motion Simulator (RMS) to generate the motion environment, as described in chapter 3 and illustrated in Figure 5-1.

Displays

Figure 5-1. RMS platform, reconfigurable cab, and experimental setup including three touch-screens.

The ride motion profile was recorded from linear and angular accelerations, collected from accelerometers centered along both the longitudinal and lateral axes of High Mobility Multi-purpose Wheeled Vehicle (HMMWV) driving across the

75

U.S. Army Research Laboratory’s Ground Vehicle Experimentation Course at Aberdeen Proving Ground (Aberdeen, MD). The six degrees of freedom acceleration signals (vertical, lateral, longitudinal, roll, pitch, and yaw) were input to the RMS to simulate the motions of the vehicle. A 60-second segment of relatively consistent acceleration data was then looped to create a continuous acceleration ride profile commanding the RMS movements (Table 5-2). Table 5-2. Acceleration data for 6DOF ride profile. Lateral, longitudinal, and vertical data are reported in m/sec2. Roll, pitch, and yaw data are reported in rad/sec 2. Channel

r.m.s.

Average

Max

Min

Crest Factor

Longitude Latitude Vertical Roll Pitch Yaw

1.130 0.806 1.360 1.122 1.384 0.555

0.144 -0.483 0.440 -0.001 0.000 0.002

4.729 1.680 7.348 5.782 5.834 2.175

-3.556 -3.904 -6.855 -4.688 -8.850 -3.083

4.183 4.846 5.404 5.154 6.394 5.556

Crest factors, which are proportions of the maximum peak acceleration divided by the root-mean-square of the acceleration signal, provide an indication of the variability of the vibration exposure. Values less than 9.0 indicate a sufficiently smooth exposure that does not have significant peak disturbances (e.g. potholes in the road), and that nominal vibration analyses will provide accurate estimates of perceived comfort and fatigue boundaries (ISO 2631-1). To further evaluate the ride profile, the running r.m.s. method shown in Equation 5-1 was used to “filter” the influence of “occasional shocks and transient vibration” (ISO 2631-1).

76

⎧⎪ 1 t0 ⎫⎪ 2 aw (t 0 ) = ⎨ ∫ [aw (t )] dt ⎬ ⎪⎩τ t −τ ⎪⎭

1 2

(5-1)

Accelerations in the Pitch rotational axis were dominant (Figure 5-2). A roll-dominant condition was created by rotating the pitch input accelerations about the vertical axis, so that the pitch motions were experienced in the Roll direction, about the longitudinal axis. Likewise, the roll accelerations of the original data were then experienced about the (negative) lateral axis. This effectively simulated the occupant seated sideways, facing the right interior side in the vehicle. The generated ride profiles are the driving inputs for the RMS cab, and do not exactly represent the vibrations experienced by the seated operators. The additional transmissibility of the HMMWV seat, though quite firm and thin, presents a source of variability that is not accounted for in the subsequent calculations of vibration effects on the human.

Figure 5-2. A power spectral density of the angular acceleration commands input to the RMS motion platform for the pitch dominated drive file.

77

Equations from the ISO 2631-1:1997 vibration standard were used to decompose the multi-axis ride motion into its axial components. The first step was to determine a frequency-weighted acceleration profile using 1/3 octave bands and weighting tables given in ISO 2631. Then the running r.m.s. method was used to incorporate the “occasional shocks and transient vibration… given as the maximum in time of aw(t0)”:

⎧⎪ 1 ⎫⎪ 2 a w (t 0 ) = ⎨ ∫ [a w (t )] dt ⎬ ⎪⎩τ to −τ ⎪⎭ to

1 2

.

(5-2)

After determining the frequency-weighted acceleration for each primary axis, the total vibration value is calculated by weighted-sum-of-squares, including multipliers for the horizontal (1.4) and vertical (1) axes:

(

).

1 2 2 2 z wz

av = k a + k a + k a 2 2 x wx

2 2 y wy

(5-3)

Finally the vibration dose value is calculated, which is a single numerical value incorporating the direction-, frequency-, and magnitude-dependent functions of 6DOF ride motion:

⎫ ⎧ 4 VDV = ⎨∫ [av (t )] dt ⎬ ⎭ ⎩0 T

1 4

(5-4)

SETUP AND MOTION SICKNESS CONTROL The RMS cab was partially enclosed by a tarp located on the cab directly in front and to the sides of the participant. This limited the participants’ view of the experimenters, but allowed for some visual reference to the external

78

environment, which was thought to limit the incidence of motion sickness. The RMS setup included a HMMWV seat with a 10° reclined seatback and headrest, steering handles in front of the subjects, and the three resistive-touch displays illustrated in Figure 5-3. The first display (“Forward”) was located directly in front of the participant with the middle of the display adjusted to the participant’s seated eye height. The second display (“Upward”) was located directly above the first display, elevated approximately 45° from the horizontal plane, and oriented perpendicular to the line-of-sight. The third display (“Lateral”) was mounted directly to the right of the seated participant, and adjusted to the participant’s seated eye height. Each display was mounted approximately 60 cm away from the seated eye position. The steering handles were located 30 cm in front of the participant, approximately 30 cm below the Forward display. An electrical switch was positioned on the right handle so that it was depressed by the middle phalanx of the right index finger when holding the hand rest. The release of the switch indicated the end of the participants’ reaction time to a target stimulus, and the initiation of the movement time.

79

Figure 5-3. The RMS configured with three touchpanel displays. The participant is illustrated reaching to the center of the Lateral display. Note that the targets visible in this picture represent the Physical targets that were affixed to the displays.

The white circles on the touch-screens (1.28 cm, 2.54 cm) depicted in Figure 5-3 represent Physical targets that were affixed to the displays. Digital targets were presented on the displays as white circles also. Electrical signals were recorded for timing of the target presentation, button release, initial contact with the touch-screen, and the location of the contact with the displays. Target presentation and data acquisition code were written in Q-Basic. Target type was defined as either Digital targets (filled-in circles) presented on the displays or Physical targets (circular discs, 0.64 cm thick) that were affixed to the front of the display. Four target sizes (diameter) were used in the experiment, and denoted alphanumerically: TS-1 (0.64 cm), TS-2 (1.27 cm), TS-3 (2.54 cm), and TS-4 (5.08 cm). Target type and target size conditions were

80

overlapped to reduce the required number of trials. The “Small”, “Medium”, and “Large” sizes for the Digital targets are denoted by D1, D2, and D3, respectively. The overlapping sizes for the Physical targets are P2, P3, and P4. Using a telephone keypad as a reference, the Digital target locations were numbered 1-9 (Figure 5-4a), while the Physical targets were in the four quadrant spaces between the Digital targets (Figure 5-4b), as well as at the center of the display. a)

b)

Figure 5-4. Target configurations: a) Digital targets b) Physical targets.

A ten-camera VICON motion capture system recorded the upper body kinematics of the subject. Reflective markers were placed on anthropometric landmarks to recreate the rigid body movements in a digital environment (Figure 55). Movement of the cab was subtracted from global fingertip movement to determine the finger’s movement with respect to the cab. These trajectory data of the fingertip were used in the subsequent reach analyses and modeling.

81

Figure 5-5. Digital human with reflective markers in the home resting posture.

EXPERIMENTAL DESIGN The experiment consisted of three reaching tasks using within-subject experimental designs with replications, which are referred to as: 1) Location by Size by Type, 2) Location by Size, and 3) Location by Type, and are described below. Table 5-3 outlines the independent measures that were evaluated within each block, and Figure 5-6 details the conditions and control factors used in each block. A balanced incomplete block design (BIBD) was used that enabled a reduction of the total number of experimental trials. Participants performed a total of approximately 640 reaches during the experiment, and were given a 30-minute break in the middle of the experiment to reduce potential vibration-induced or taskrelated fatigue. In the ride motion conditions, participants experienced up to 180 seconds of continuous ride motion before a 30-second rest between trials.

82

Table 5-3. Overlapping conditions of experimental design. Independent Variables

Block 1

Ride Motion

Block 2

Block 3

X

X X

Target Location

X

X

Target Size

X

X

Target Type

X

X

Location by Size by Type

Location by Size

Block 1: 10 * 3 * 3 * 2 = 180 reaches

Block 2: 9 * 3 * 3 * 3 = 243 reaches

Experiment conditions: 1. Reach Direction (3) 2. Target Size (3) 3. Target Type (2)

Experiment conditions: 1. Reach Direction (3) 2. Target Size (3) 3. Ride Motion (3)

Control factor: ride motion (Stationary)

Control factor: target type (Physical)

Location by Type Block 3: 12 * 3 * 2 * 3 = 216 reaches Experiment conditions: 1. Reach Direction (3) 2. Target Type (2) 3. Ride Motion (3) Control factor: target size (2.54 cm) Figure 5-6. Experimental block design with overlapping conditions, including control factors in each block.

REACHING TASKS Each task consisted of a series of trials in which the participants reached and pressed the target “as quickly and accurately as possible”. Participants were required to respond to the simultaneous presentation of a visual target and audible tone by quickly reaching out to press a white circular target with the right index fingertip. Three frequencies were used for the signal tone, depending on which display the target was presented. The frequencies of the tones for targets

83

presented on the Forward, Lateral, and Upward displays were 261.6 Hz (“middle C”), 523 Hz, and 1046 Hz respectively, and sounded for 1/16th of a second. Musically, each of theses tones were “C” notes, each separated by one octave. An additional tone was generated in response to a successful target press registered by the resistive-touch display. This tone (800 Hz for 0.27 sec) sounded approximately 0.08 sec after the initial contact, due to a processing delay for the calculation of the contact location. After successfully completing the reach, the participant immediately returned the right hand to the hand rest and waited for the next trial to begin, after a random delay included mitigating anticipatory effects. The tasks were part of a within-subjects experimental design that collectively assessed the effects of three motion conditions (Roll, Pitch, and Stationary) and three target characteristics (Location, Size, and Type) on reach speed and accuracy. The Location by Size by Type block was performed in the stationary condition only, where participants performed rapid reaches to each of the three displays (Forward, Lateral, or Upward). Each target size and target type conditions were presented. The order of the motion conditions was randomized across participants and the target sizes, target types, and target locations were randomized within each condition. This block formed a 27-cell design of Location x Size x Type, consisting of 180 trials for each participant. The Location by Size block was performed under the three motion conditions (Roll, Pitch, and Stationary), consisting of reaches to Physical targets located on each of the three displays. Within each motion condition, participants

84

reached to the P2, P3, and P4 target sizes. The order of the motion conditions was randomized across participants, while the target sizes and locations were randomized within each condition. This block formed a 27-cell design of Motion x Location x Size, where each participant performed 243 trials under these conditions. The Location by Type block also used the three motion conditions, where participants were presented both target types on each of the three displays. All targets were 1” circles. The order of the motion conditions was randomized across participants and the target types and touch-screen locations were randomized within each condition. This block formed a 27-cell design of Motion x Location x Type. There were 216 trials performed by each participant under these conditions. EXPERIMENTAL PROCEDURE Upon arriving at the simulator laboratory, all participants were briefed about the purpose of the study and experimental procedures, were introduced to the equipment and witnessed its operation, and completed the human use consent forms. Participants were seated in the RMS cab and restrained only with a lap belt, allowing for unrestrained torso movement. Participants were given practice trials of the reaching tasks to familiarize themselves with the experimental setup, and were instructed to practice the tasks until the experimenter observed that participants generally followed the instructions to move as quickly as possible while accurately touching the target locations with their right index fingertip.

85

Participants then completed each of the three tasks in a random order with a rest break given between the second and third tasks. Prior to beginning each task, the participants completed an additional block of 20 practice trials specific to that task. Each task was divided into blocks of approximately 30 reaches, requiring up to 180 seconds to complete each block. A rest period of 30 seconds was provided between each block. After two of the four tasks were completed, participants exited the RMS, took a 30-minute break, and then returned to the RMS to complete the two remaining tasks. In total, the participants were in the RMS for approximately 65 minutes of testing, of which 35 minutes were under ride motion perturbation. DATA REDUCTION Despite the practice sessions, an order or learning effect was significant (ANOVA, p < 0.05) as subsequent reaches yielded increased accuracy with respect to reach order. To remove this effect, the first reach was removed from the analysis dataset. Reaches were removed successively from the beginning until an order effect was no longer significant. All of the empirical data were subjected to diagnostic analysis to eliminate outlier data with excessive leverage or residual error. “Leverage” in this case can be viewed as the normalized effect that a data point has on the least squares fit of the data. Faraway (2004) suggests that the leverage for a data point should not be larger than twice the mean leverage for the dataset, and should be removed. Data points may have large residual error, the difference between the actual point and

86

the mean of the dataset, but are not considered outliers if the residuals are normally distributed and have constant variance throughout the dataset. Accuracy of the reach endpoint (radial distance) was measured as the distance from the center of the target to the center of pressure on the touchscreen. To illustrate the operational consequences of reach accuracy, a cumulative histogram distribution was used to evaluate the radial distances, such that percentages of reaches within 1”, 1.25”, 1.5”, and 1.75” of the target center were computed. Radial distances were also evaluated in a continuous dataset of effective target widths, interpreted here as the minimum target size that would have been required to ensure 96% (mean ± 2 SDs) of the actual endpoint fingertip locations. Statistical models were represented by mixed linear models in JMP® 12.0 (SPSS, Inc., Chicago) and R-2.0.1 for Windows (Ihaka and Gentleman, 1996). As relevant for each task, models were generated for reaction time, movement time, and radial distance, based on subjects, ride motion, display location, target size, target type, and the associated interactions. Analysis of variance with repeated measures (ANOVAR) was used to assess significance at the 0.05 alpha-level, unless otherwise indicated. The standard deviations and standard errors are provided, along with the mean values for the dependent measures, to show the variability in the dataset (standard deviation), while at the same time illustrating the precision of the expected mean values (standard error) of the distribution. These three values are reported below for each dependent measure in the following format: (mean (standard error) ± standard deviation).

87

Results Ride

motion

was

found

to

significantly

degrade

manual

reach

performance, and discussed in detail below with respect to reaction times (Table 5-4), movement times (Table 5-5), and radial distance of the fingertip with respect to the target center (Table 5-6). While both motion conditions (Roll and Pitch) showed significant effects on reaction times, movement times, and endpoint accuracy of reaches performed in all reaching tasks, results between the Roll and Pitch motion conditions were not significant (Figure 5-7). Hence, the motion conditions are combined and are collectively defined and referred to only as “ride motion”. Individual motion results are provided in some cases to show relevant trends and noteworthy effects.

Comparison of motion conditions 400

20

Reaction Time (ms)

16

350 12 325 8

300

4

275 250

0 Static

Roll

Pitch

Target Width (mm)

375

Motion Condition Figure 5-7. Increases in reaction time and effective target width due to ride motion (across all reaching tasks).

Hypothesis 5.1: Ride motion will result in longer reaction times. Due to the instability of the environment, ride motions significantly affect the vehicle operator’s ability to respond quickly to target stimuli. Table 5-4 shows

88

that reaction times (RT) were significantly affected by several factors, including ride motion, display location, and target size, as well as some of their interactions. RT significantly increased from the Stationary condition (312 (1.5) ± 112 ms) to the Roll condition (353 (3) ± 163 ms), and to the Pitch condition (356 (3) ± 164 ms). There also was a significant interaction between Ride Motion and Display Location. In the Stationary condition, RT were significantly longer for reaches to the Upward display (339 (2.5) ± 105 ms) than reaches to the Lateral display (317 (2.4) ± 106 ms), while the reaches to the Lateral display also were significantly longer than reaches to the Forward display (294 (2.4) ± 105 ms). Under the ride motion conditions, RT for reaches to the Forward display (307 (4.1) ± 181 ms) were significantly shorter than to the Lateral (358 (4) ± 178 ms) and Upward displays (367 (4.1) ± 178 ms); however there was not a significant difference in RT for reaches to the Lateral and Upward displays. Table 5-4. ANOVA results for reaction times from combined results.

Source

N

DF

S.of S.

F Ratio

Prob > F

Ride Motion (RM)

2

2

1.47

53.0

< 0.0001

Display Location (DL)

2

2

3.16

114

< 0.0001

Target Type (TT)

1

1

0.05

3.44

0.084

Target Size (TS)

2

2

2.00

71.8

< 0.0001

RM * DL

4

4

0.23

4.22

0.002

RM * TT

2

2

0.08

2.83

0.048

RM * TS

4

4

0.30

5.37

0.000

DL * TS

4

4

0.14

2.58

0.036

TT * TS

2

2

0.47

16.8

< 0.0001

89

Hypothesis 5.2: Reaction times will be significantly affected by the display location for three-dimensional reaching tasks. Recall that subjects were instructed to look forward until hearing the auditory stimulus of the target presentation, resulting in the significant effect of Display Location shown above in Table 5-4. RT were significantly shorter to the Forward display (304 (2.3) ± 142 ms), than to each of the other displays. Similarly, RT were significantly shorter for reaches to the Lateral display (340 (2.2) ± 142 ms) than to the Upward display (357 (2.3) ± 139 ms). There also was a significant interaction between Display Location and Target Size. In each Display Location condition, the RT in response to Small targets (Forward: 318 (2.2) ± 102 ms; Lateral: 363 (2.2) ± 104 ms; Upward: 377 (2.3) ± 102 ms) were significantly longer than RT in response to Large targets (Forward: 285 (4.7) ± 132 ms; Lateral: 327 (4.6) ± 130 ms: Upward: 343 (4.6) ± 131 ms). Moreover, RT for Medium targets were not found to be significantly different than RT in response to either Small or Large targets. ADDITIONAL REACTION TIME ANALYSES (POST-HOC) Target size also influenced reaction times, significantly decreasing with each increase in target size: Small (349 (1.4) ± 107 ms), Medium (323 (2.8) ± 152 ms), and Large (307 (3.6) ± 176 ms). The interaction between Ride Motion and Target Size also was significant. In the Stationary condition, the only significant difference in RT was observed between the Small target (326 (1.9) ± 102 ms) and the Large target (303 (2.7) ± 107 ms), while RT in response to a Medium target were not significantly different from the RT in response to either

90

the Small or Large targets. Under the ride motion conditions, RT in response to the Small target (365 (2.5) ± 148 ms) were significantly longer than RT to the Medium targets (343 (4.9) ± 176 ms) and Large targets (326 (6.0) ± 193 ms); however RT to the Medium targets were not significantly different from the RT to the Large target. The final interaction significantly affecting RT was between target size and target type, even though target type was not a significant main effect. For the Digital targets, RT in response to Small targets (356 (2.1) ± 104 ms) were significantly longer than to either the Medium targets (331 (5.0) ± 123 ms) or Large targets (322 (6.5) ± 134 ms); however there was not a significant difference observed in RT for reaches to the Medium and Large targets. With respect to the Physical targets, the RT in response to the Large targets (314 (2.3) ± 103 ms) were significantly shorter than to the Medium targets (342 (2.2) ± 104 ms) or Small targets (349 (1.2) ± 102 ms), but there was not a significant difference between the Medium and Small targets. Hypothesis 5.3: Ride motion will result in longer movement times under vibration than in the stationary condition. Results presented in chapter 4 showed that ride motion perturbation resulted in longer movement times with respect to temporally-unconstrained reaches that only emphasized accuracy. Likewise Table 5-5 shows that ride motion significantly affects movement times; however movement times for these temporally-constrained rapid reaches were shorter under ride motion (444 (3.3) ± 103 ms), than reaches performed in the Stationary condition (460 (1.6) ± 96 ms),

91

as opposed to the results for temporally-unconstrained movements. While the effect is significant, the magnitude of the difference is small (3.6%). However the negligible effect in movement time can be reconciled by considering the significantly degraded accuracy under ride motion, reported under Hypothesis 5.5. Table 5-5. ANOVA results for normalized movement times from combined results.

Source

N

DF

S.of S.

F Ratio

Prob > F

Ride Motion (RM)

2

2

0.201

2.69

0.068

Target Size (TS)

2

2

0.004

0.05

0.947

Target Type (TT)

1

1

0.017

0.45

0.500

RM * TS

4

4

0.453

3.03

0.017

TS * TT

2

2

0.631

8.44

0.001

Hypothesis 5.4: Movement times to digitally presented targets will be shorter than the movement times to physical pushbuttons for threedimensional reaching tasks. The analysis of variance shown in Table 5-5 shows that Target Type did not have a significant effect on movement times. However, recall that use of overlapping conditions of target size and target type resulted in singularities for the Small Digital target and the Large Physical target, as shown in Figure 5-8. Across vibration conditions, mean MT for each target type were not significantly different, as may be seen by the similar inverted-U shapes of their respective distributions in Figure 5-8. Likewise Target Size did not have a significant effect, whether modeled as a nominal factor (levels: Small, Medium, Large), or as a continuous factor using the target diameters. In both cases, the effect of Target Size on mean MT were not statistically significant, despite the disparity at the 12.7 cm and 25.4 cm target diameters between target types.

92

There was a highly significant interaction between Target Size and Target Type that reconciles this disparity. Reaches to the 12.7 cm target exhibited significantly shorter MT when the target was Physical (427 ± 98 ms) than when the target was Digital (492 ± 88 ms). The opposite was observed for reaches to the 25.4 cm target, where MT were shorter to the Digital targets (461 ± 92 ms) than reaches to the Physical targets (482 ± 104 ms). The percentage of successful reaches (hit rate) is also plotted on Figure 5-8, which declines rapidly as the target size decreased.

Effect of Target Size and Target Type on MT Movement Time (Physical) Hit Rate (Physical)

525

100%

500

80%

475

60%

450

40%

425

20%

400

Hit Rate (%)

Movement time (ms)

Movement Time (Digital) Hit Rate (Digital)

0% 6.35 (0.25)

12.7 (0.5)

25.4 (1.0)

50.8 (2.0)

Target Size - mm (in) Figure 5-8. Effects of target size and target type on the timing and accuracy of reaching tasks.

ADDITIONAL MOVEMENT TIME ANALYSES (POST-HOC) There also was a significant interaction between Ride Motion and Target Size, despite that Target Size was not a significant main effect. In the stationary

93

environment, reaches to the Large target (447 (3.2) ± 119 ms) were significantly shorter than reaches to the Small targets (468 (2.1) ± 113 ms) and Medium targets (473 (3.0) ± 118 ms). Under ride motion, reaches to the Large target (431 (6.6) ± 119 ms) were significantly shorter than reaches to the Medium targets (490 (3.0) ± 118 ms), as well as reaches to the Small targets (466 (2.8) ± 113 ms). However, contrary to Fitts’ law, reaches to the Small targets were significantly shorter than reaches to the Medium targets. This effect is further discussed in combination with the effects of target size on accuracy under the following hypothesis, and in the Discussion section. Hypothesis 5.5: Variability of reach end points will be larger under ride motion than stationary condition. Table 5-6 shows that ride motion significantly affected the accuracy (mean ± standard error) of temporally-constrained reaching tasks. Across all conditions, the endpoint variability, given in terms of the radial deviation (RD) of the reach endpoint to the target center, of reaches performed in the Stationary condition (6.1 ± 0.1) was significantly less than the RD of reaches performed in the Ride Motion condition (9.5 ± 0.2). This 56% increase in spatial error is inversely proportional to the significant (though minimal) decrease in movement times under ride motion that was discussed earlier. Ride motion also had significant interactions with the Display Location, Target Type, and Target Size. With regard to Display Location, RD was not significantly different for reaches to any of the displays in the Stationary condition. However, under ride motion, reaches to the Lateral display (8.6 ± 0.3) were significantly more

94

accurate than reaches to the Upward display (10.3 ± 0.3). Conversely for the Target Type, there was not a significant difference in RD observed under ride motion; however in the Stationary condition, reaches were significantly more accurate to the Physical targets (5.8 ± 0.1), compared to the Digital targets (7.0 ± 0.2). There was only one significant interaction between Ride Motion and Target Size on RD, in the Stationary condition, where reaches to the Small target (6.0 ± 0.1) were significantly more accurate than reaches to the Large target (7.0 ± 0.2). Table 5-6. ANOVA results for radial distances from combined results.

Source

N

DF

S.of S.

F Ratio

Prob > F

Ride Motion (RM)

2

2

7,216

70.6

< 0.0001

Display Location (DL)

2

2

609

5.96

0.003

Target Type (TT)

1

1

213

4.2

0.041

Target Size (TS)

3

3

5,209

34.0

< 0.0001

RM * DL

4

4

566

2.77

0.026

RM * TT

2

2

2,001

19.6

< 0.0001

RM * TS

6

6

1,600

5.22

< 0.0001

DL * TT

2

2

117

1.15

0.318

DL * TS

6

6

574

1.87

0.082

RM * DL * TT

4

4

537

2.63

0.033

Figure 5-9 shows the effect of ride motion on the hit rate of rapid reaching tasks. The solid, light horizontal line indicates that 95% of the reaches in the Stationary condition fell within 1.25” of target center, while in the motion conditions, 95% fell within approximately 1.75”. The 0.5” difference between the Stationary and moving conditions was also observed at the 90% accuracy line.

95

Hit Rate (%)

Distance from target center Figure 5-9. Effect of ride motion on reach accuracy, as a function of the distance from the reach endpoint and the center of the target.

Hypothesis 5.6: Reaches will be more accurate to physical pushbuttons than reaches to digitally presented targets. Target Type was shown in Table 5-6 to be a significant factor affecting the endpoint accuracy. Across motion conditions, reaches to Digital targets (7.7 ± 0.3) were 13% more accurate (p < 0.05) than reaches to Physical targets (8.7 ± 0.1). There was also a significant three-way interaction between Target Type, Ride Motion, and Display Location, despite that the two-way interaction of Target Type and Display Location was not significant. Figure 5-10 illustrates the differences between the three conditions and their factor levels. Reaches to Physical targets were significantly more accurate than reaches to Digital targets for the Forward and Vertical displays in the Stationary condition, while there were no significant differences observed under Ride Motion, across displays and target types.

96

Digital targets

Physical targets

Radial deviation (mm)

25 20 15 10 5 0 Forward

Lateral

Vertical

Stationary

Forward

Lateral

Vertical

Ride Motion

Figure 5-10. Interactions of ride motion, display location, and target type.

Discussion Whole-body ride motion contributed to approximately 13.5% longer reaction times to stimulus presentation, 3.6% shorter movement times, and a 56% increase in endpoint variability. With respect to reaction time (RT), ride motion, display location, and target size all significantly affected RT, as well as two-way interactions between them. Increases in reaction times are consistent with increased variability in the movement planning process due to ride motion, where the vehicle motion introduces unpredictable orientations of the participant, relative to intended targets. Performing reaching tasks in this dynamic environment requires a highly adaptive movement planning process. In addition to vibration-induced movement alterations, the interaction between ride motion and target size may relate to visual acuity issues, where the three-dimensional position of smaller targets may be more difficult to ascertain under ride motion.

97

This effect of the ride motion and target size interaction may translate to the significant increases relating to movement time as well. Movement times also were significantly affected by ride motion, and the two-way interaction between target size and target type. Figure 5-8 illustrated this interaction through the similar inverted-U shape distributions exhibited for MT. It is interesting to note that the distributions appear shifted by one target size, where for example, there was not a significant difference between movement times for the 12.7 cm Digital target and the 25.4 cm Physical target. In addition, there is an inverse relationship in mean movement times for reaches to the 12.7 cm and 25.4 targets. At the 12.7 cm target size, reaches are significantly quicker when the target is Physical, whereas the opposite is true for reaches to the 25.4 cm target. More importantly, the hit rates were not statistically different for either target size or target type. Vehicle ride motions were shown in chapter 4 to result in longer movement times for temporally-unconstrained (self-paced) reaches, where accuracy is the primary goal of the push-button reaching task. However, when significant temporal constraints were introduced, the ride motion perturbation resulted in a slight decrease (3.6%) in movement times. While the difference in movement times may be negligible, the accuracy of reaches in the Ride Motion condition was severely degraded, compared to the reaches performed in the Stationary condition. Despite instructing the subjects to be both fast AND accurate, it appears that the most common movement strategy was to maximize speed, at a cost of decreased accuracy. This may partially due to an inability to

98

be highly accurate in such a dynamic environment, relegating the participants’ movement control to only movement speed. The related manual coordination and feedback control mechanisms are investigated and further discussed in Chapter 6. Figure 5-9 showed that a circle encompassing 90% of the reaches would have to be 0.5” larger during the ride motions experienced in this study, in order to expect endpoint accuracy under ride motion equivalent to that observed in a stationary environment. Similarly, a circle encompassing 95% of the data would also require a 0.5” increase to result in equivalent reach accuracy. This severe degradation in reach accuracy due to the ride motion environment strongly supports the need for improved displays and target designs, as well as mitigation strategies to attenuate ride motion effects in ground-based vehicles. The utility of touchpanel displays seem to be highly dependent on the nature of the task, specifically on whether speed or accuracy is more important. When the priority of a reaching task is endpoint accuracy, and assuming that speed is not essential, the task would most benefit from a physical target, rather than a digital presentation on a display. However, if the task is temporally unconstrained, digital touchpanel displays can provide additional functionality through hierarchical menus and button sizes that can be varied in presentation to assist in more critical tasks. One method for activating targets on a touchpanel display is through use of a “lift-off” strategy (as opposed to the “land-on” strategy used in this study), in which the button is not selected until the finger is removed from the screen. This allows the user to touch the display, and use contact friction to adjust the fingertip placement to the desired position before selecting

99

the target by removing contact with the display. Displays could also be designed to adjust images based on vehicle movements, possibly enlarging menus and buttons as a function of the ride motion environment. Regardless, vehicle design must emphasize mitigating ride motion effects from a system-level view of the transmissibility of terrain-induced ride motion through the tires, wheels, suspension, and seating (e.g. active suspensions tuned to reduce the power in the 4-6 Hz frequency band). As was the recommendation regarding the use of touchpanel displays, design recommendations of control size remains dependent on the desired functionality of the task, particularly whether speed or accuracy is of greater concern. Expected performance levels are still predominantly dependent on the occupants’ manual dexterity. If population-based manual coordination levels were available, control and display designs could be designed to accommodate variations in skill and motivation. Another intersubject factor, and sometimes intrasubject, that affects the speed and accuracy of pointing tasks is the subject’s motivation. For the reaches performed in the present study, subjects were instructed to be both fast and accurate, though the speed-accuracy tradeoff suggests that this is not possible. The Shannon formulation of Fitts’ law (Equation 5-4) clearly indicates that the movement amplitude (A) and effective target width (We) are inversely proportional.

⎛ A ⎞ MT = a +b × log 2 ⎜⎜ + 1⎟⎟ ⎝ We ⎠

100

(5-4)

As discussed previously in Chapter 4, MacKenzie and Buxton (1992) and later Murata and Iwase (2001) have attempted to modify the Fitts’ law formulation to predict movement times for rapid 2D and 3D pointing tasks, respectively. The latter formulation attempted to incorporate the direction of movement:

⎤ ⎡ ⎛A ⎞ MT = a +b × ⎢log 2 ⎜ + 1⎟ + c × sinθ ⎥ ⎝W ⎠ ⎦ ⎣

(5-5)

This formulation separates movement amplitude from elevation, and does not include the elevation component in the logarithmic function. Despite that the Murata formulation does not seem to account well for reaches to targets that are distant in both horizontal and vertical directions, it is an adequate foundation to include variables for motivation and ride motion effects. The previous and current studies suggest that an additional parameter could be included in the Fitts’ formulation to increase the robustness to include reaches in a dynamic environment. In the case of ride motion perturbations on movement planning and feedback control, equations from the ISO 2631 standard can be used to generate the vibration dose values (VDV), a single numeric value that incorporates the direction-, frequency-, and magnitude-weighted effects of vehicle vibration. Ride motion has been shown to affect the reach difficulty (i.e. index of difficulty, ID), therefore the VDV would be a multiplier of the entire ID. It has been established that an unpredictable motion environment will degrade reaching movements. Reaches performed in a stationary environment were initiated more quickly and were more accurate than reaches performed under vehicle motion dominated by either roll or pitch accelerations, as described in chapter 3. These findings are consistent with the theories of movement 101

planning and control, described in chapter 2, that suggest that the ability to predict the environment is essential to the movement planning and control processes (Wolpert et al. 1995; Wolpert et al. 2000; Erlhagen et al. 2002). The ability to quickly and accurately make reaching movements will be critical in the future in-vehicle environments proposed for the U.S. Army. This study has indicated that when a person is exposed to moderate vehicle motions, they will take longer to initiate the motion and move more quickly, which will result in 30-50% less accuracy. The nature and extent of the degraded accuracy, as well as how visual and proprioceptive feedbacks are affected by ride motion are discussed in Chapter 6.

102

Chapter 6 Use of visual and somatosensory feedbacks under ride motion This chapter examines the ability to perform rapid pointing tasks under ride motion, with and without vision. The goal of this study is to explore the availability and utility of visual and proprioceptive feedbacks in the manual control of reach trajectories perturbed by ride motion.

Introduction The central nervous system (CNS) is assumed to serve as a master controller, where movements are planned and subsequently executed (Bernstein 1967; Schmidt and Lee 2005) using visual and proprioceptive feedbacks (van Beers and Sittig 1996) and/or efference copy of the muscular command (see Kelso 1982 for review). As a target-directed movement is begun, the initial frame of reference is body-centered, or egocentric, and largely determined through somatosensory feedback (Soechting and Flanders 1989, 1992; Ghez 1991; Riemann and Lephart 2002). As the fingertip approaches the target, vision becomes the dominant feedback mechanism as the CNS transitions to an exocentric reference frame, where necessary corrections can be made to ensure successful completion of the reach (Gordon et al. 1994; Paillard 1996). The increase in endpoint variability in absence of visual feedback is known to increase endpoint variability, and supported by the work of Keele and Posner (1968) and Zelaznik et al. (1983). In other words, these results suggest that the 103

CNS differentially uses the most informative feedback available in various stages of a movement. It is believed that removing visual feedback prevents the transition to an exocentric reference frame, and the remaining afferent feedback mechanisms do not have sufficient resolution to make compensatory movements during rapid reaching. This study explored several planning and movement strategies during rapid reaches with and without visual feedback to touch-screens under both stationary and ride motion environments. Delays to movement onset were imposed after visual feedback was removed to investigate the fidelity of the internal representation/model in short-term memory of the target location, if it exists. Elliot and Madalena (1987) noticed that reach accuracy continued to degrade 2, 5, and 10 seconds after vision was removed. Adamovich et al. (1998) investigated the reach accuracy associated with how the target was presented to the subject, visually or kinesthetically (where the arm was moved into the correct posture, but vision was not available). Their results were largely inconclusive, other than confirming that people can use proprioceptive feedback to perform visually-occluded tasks, albeit with degraded accuracy. Rosenbaum et al. (1999) attempted to delineate between Adamovich’s dichotomy, by having subjects perform the reaches with the opposite arm, as he argues that the neuromuscular system may be able to translate proprioceptive joint information from one arm to the other. While Rosenbaum had some success in demonstrating a posturebased guidance, he also admitted that the brain may simply be attempting to replay the same motor command. Finally in either case, Westwood et al. (2003)

104

argues that if memory-guided movements can be used to accurately perform visually-occluded tasks, then a sufficiently brief interval must exist where visuallyoccluded reaching is as accurate as visually-guided movements. Each of these studies used the delay time as a discrete independent variable. A source of considerable noise in these paradigms is the variability in reaction time, where a long reaction time to a stimulus lengthens the total response time to target contact, thus requiring additional time that the mental image must be stored. The experimental design used in this study uses a similar binning strategy for independent delay conditions, but incorporates the reaction time in the post-hoc analysis. One issue that remains unclear is whether the CNS makes discrete corrections or employs continuous control to maintain sufficiently accurate endeffector trajectories. In feed-forward, or open-loop movements, optimal end effector trajectories “theoretically” exhibit temporally symmetrical speed profiles (Morasso 1981), whose peak tangential velocities and movement times are proportional to the distance of the reach (Newell et al. 1984). In feedbackcontrolled reaching, the profiles exhibit temporal asymmetry (e.g. Uno et al. 1989; Elliott et al. 1999; Todorov 2004), with motion velocities skewed to the right in the decelerative phase. This skewness, sometimes accompanied by additional local extrema in the speed profile, is generally considered evidence of feedbackbased corrective submovements (Crossman and Goodeve 1983; Meyer et al. 1990).

105

Schaal (2002) presents an alternative explanation to the preceding where three-dimensional hand movements are comprised of successive planar movements, referred to as segmented control. Technically, this segmentation can be executed in a strictly feedforward manner if the movement requires two planar curves to travel from origin to destination, thus the skewness may not be inherently tied to feedback control. Figure 6-1 illustrates this hypothesis using the superposition of negative cosine curves, similar to existing theories of coordinated movement control (Adamovich et al. 1984, 1994; Flash 1990; Flanagan et al. 1998; Sanger 2000). The superposition of these separate motor commands results in the observed skewed bell-shaped speed profile (Figure 6-1b). a)

Speed

Speed

b)

Time

Time

Figure 6-1. Speed profiles: a) Subsequent discrete optimal movements, b) Movement resulting from superposition of planar movements (Rider et al. 2004).

Early work by Keele and Posner (1968) suggest that the lack of early corrections to the original movement is due to the fact that visual information takes time to process. Visual feedback loops have been estimated to be between 85 and 135 ms (Soechting et al. 1989; Carlton 1992), whereas proprioceptive loops are argued to be as long as 150 ms (Jeannerod 1991). Modifications to the trajectory after 150 ms have elapsed presumably include feedback information, regardless on whether changes in the movement are discernable.

106

Conversely, Feldman et al. (1994) suggests that control processes that govern rapid movements end before the fingertip reaches peak velocity. Therefore if visual feedback is not available and proprioceptive feedback cannot provide both timely and informative feedback, the CNS may not detect trajectory deviations and will therefore continue executing its original motor commands without modification. If the movement indeed remains open-loop, this should result in a high correlation between the direction of each movement’s deviation (radial deviation) at the peak velocity and its radial deviation at the target, since corrections could not be made. Messier and Kalaska (1999) and Heath (2005) have investigated these correlations in temporally-unconstrained reaching and confirmed that feedback processes appear to be utilized in compensating for early variability in these tasks. The work reported here explores this concept. MOVEMENT PLANNING Bernstein (1967) presented seminal theories on motor coordination and the planning and execution of discrete reaching tasks, including online trajectory modifications based on visual and somatosensory feedback. The visuomotor system compares visual discrepancies during reaching tasks between the relative end-effector position and the target in order to make necessary movement corrections (see Desmurget et al. 1998 for review). Some studies also have investigated whether the CNS retains a representation of target location, even after visual feedback is removed (e.g. Soechting and Flanders 1989; Berkenblit et al. 1995; Sergio and Scott 1998, Heath 2005). Elliott and Madalena (1987) agreed with the premise of memory-guided reaching, and hypothesized

107

that the visuomotor system retains some degree of target representation, even after visual occlusion. However, Westwood et al. (2003) suggest that insufficient evidence has been provided to date to support that conclusion. They argue that proof of a latent target representation would require evidence that visuallyoccluded movements are as accurate as visually-guided movements for a “sufficiently brief delay interval”. They further presume that this visuomotor processing happens in real time, since visually-occluded reaches exhibit much higher endpoint variability than visually-guided reaches. It has been suggested that an efferent copy of visuomotor responses could allow the CNS to predict future states and make preemptive corrections (Goodale 1988). Another possibility is that the CNS is able to separately utilize central vision (fast response) and peripheral vision (slow response) to enable rapid feedback control processes (Blouin et al. 1993). The issues can be simply stated: if vision is dominant, what role does proprioception have? Several studies have investigated the use of proprioceptive feedback, revealing degraded performance and variant movement strategies in deafferented individuals (e.g. Gentilucci et al. 1994; Ghez et al. 1995; Messier et al. 2003). A combination of vision- and proprioception-based studies suggests that optimal execution of rapid reaching requires both visual and somatosensory feedback. For example, Rossetti et al. (1995) investigated the dual use of visual and proprioceptive feedback mechanisms in pointing tasks, and presented evidence for a weighted combination of vision and proprioception, but did not suggest how the CNS might assign its weightings, or if the weightings might

108

change during a reach. Sarlegna et al. (2004) agreed with a weighted combination of visual and proprioceptive feedbacks, and assigned to them 45% and 55% respectively, based on linear regression models fitting endpoint deviations of the hand to unperceived shifts in the actual target location. They acknowledge that the weightings may be task-specific, but give no specific indication how the weightings would be affected by changes in task parameters. Compounding the dynamic problem of movement execution is the existence of errors in movement planning. These errors are due in part to the separate and inexact estimation of the distance and direction of the target (Messier and Kalaska 1997; Servos 2000). Additionally, physiological noise in the neuromuscular system induces errors in the executed movement, preventing the planned trajectory from being perfectly executed (Schmidt et al. 1978, 1979; Meyer et al. 1982). It also has been shown that stimulation of the neuromuscular system by whole-body vibration increases this noise component to further distort the resulting movement from the intended movement plan (Gauthier et al. 1981; Martin et al. 1991; Gribble et al. 2003). For the present study it is proposed that despite this natural and environmental variability, the mean trajectory path and mean endpoint locations of the fingertip in repeated rapid target-directed reaches is a close approximation of the originally intended movement (Soechting and Lacquaniti 1981, Faraway 2000), and can be used as a reference to compare movement variability, and the effects of vibration perturbation.

109

DYNAMIC ENVIRONMENTS Degraded manual performance under whole body vibration has been documented in several studies. Reduction in cognitive and fine motor performance was observed during, and immediately after, whole body vibration (Gauthier et al. 1981; Martin et al. 1997; Cowings et al. 1999). Further, the amount of vibration and the duration of the exposure have been shown to influence the amount of cognitive performance decrement (Schipani et al. 1998). However, the particular conditions under which motion environments degrade specific manual task performance are not completely known. That is, the mechanism of the degradation will depend on the specifics of both the motion environment and of the visuomotor task being performed. Difficulty of reaching movements in moving environments is associated with a decreased ability to predict future movement perturbations, and a mismatch between a vehicle occupants’ sensory perception and the actual environment (Riemann and Lephart 2002). Motor planning and control theories suggest that both of these factors can lead to degradations in reach performance (Wolpert et al. 1995; Wolpert and Ghahramani 2000; Erlhagen and Schöner 2002). Specifically the decreased ability or inability to predict environmental conditions will require additional time for movement planning, as well as degraded manual control of the fingertip trajectory. That is, it is hypothesized that the ride motion environment will be associated with significantly longer reaction times, longer movement times, and decreases in spatial accuracy of a manual pointing task than reaches performed in a stationary environment.

110

Materials and methods PARTICIPANTS Ten volunteers (6 men and 4 women) participated in this experiment, and were free from any knowledge or symptoms of recent cold, flu, or anti-motion sickness medications. Nine participants had experience on the Ride Motion Simulator (RMS) for 8.4 ± 5.1 hrs. Summary characteristics are provided in Table 6-7. Table 6-7. Participant summary: Stature (cm), Weight (kg), Age (yr). Male (n = 6) Mean SD Range

Female (n = 4)

Stature

Weight

Age

Stature

Weight

Age

178.4

79.0

29.9

170.2

66.6

26.2

4.4

3.9

7.0

1.3

3.1

12.3

173-185

78-87

26-45

167-171

71-77

19-46

EXPERIMENTAL SETUP This study utilized the RMS, described earlier in chapter 3, to generate the ride motion environment. The 60-second, six degree of freedom ride motion profile that was described in chapter 5 was used here to again simulate the motions of a High Mobility Multi-purpose Wheeled Vehicle (HMMWV). The motion conditions used were the same as those used in chapter 5: pitchdominant and roll-dominant. Reaching movements were performed in the RMS cab, with and without ride motion. The RMS cab was partially enclosed by a tarp placed in front and to the sides of the seat, which limited the participants’ view but allowed for some visual reference to the external environment. Hand rests were placed directly in front of the participant; the right hand rest was the global reference (0 cm, 0 cm, 0 cm – positive right, forward, and up respectively in the

111

following descriptions). A “Home” switch was placed on the hand rest and depressed by the right index finger. Three resistive-touch displays were mounted to the cab, to which subjects performed rapid pointing movements to white circular targets presented on the display (Figure 6-2).

Figure 6-2. The RMS configured with three touch-screens. The participant is illustrated reaching to a target on the Lateral touch-screen.

The displays were oriented approximately orthogonal to the visual line-ofsight, and provided the (x, y) screen coordinates of the fingertip contact to determine the endpoint location for each reach. The display locations are given in displacement (cm) with respect to the Home position of the right hand on the steering wheel. of lateral, fore-aft, and vertical The Forward display was located directly in front of the participant, vertically adjusted to place the center of the screen at eye level (display center (-10 cm, 25 cm, 30 cm), approximately 35 cm from hand rest). The Lateral display was mounted to the right side, and also

112

vertically adjusted to place the center of the screen at eye level (display center (60 cm, 25 cm, 30 cm), approximately 70 cm from hand rest). The Upward display was rigidly mounted above the front display (display center (-10 cm, 80 cm, 5 cm), approximately 80 cm from hand rest), elevated approximately 45° from horizontal. Each display was located approximately 60 cm away from the subject’s nasion. A ten-camera VICON 524 motion capture system, sampling data at 60 Hz, recorded the upper body kinematics of the participant. As with the previous studies, reflective markers were placed on essential anthropometric landmarks to recreate the rigid body movements in a digital environment. Kinematics of the fingertip (i.e. time and magnitude of peak tangential velocity) were used to evaluate reach trajectories. Reaction time, movement time, and endpoint accuracy were also recorded. EXPERIMENTAL PROCEDURE All participants were briefed about the purpose of the study and the experimental procedures, and then completed a human use consent form as part of both the US Army and University Institutional Review Board procedures. Participants were seated in the RMS cab and restrained using only a lap belt, allowing for unrestrained torso movement. Based on the analyses presented in chapter 5, the Pitch and Roll motion conditions were combined into a single motion condition; results were compared between the Ride Motion and Stationary conditions. Participants were instructed to complete reaches with the right index fingertip to visually presented circular targets on the touchscreens as

113

fast and as accurately as possible. After concluding each reach, the participants moved their hand back to the same initial position for the next trial. Practice trials to all targets were given to allow familiarization with task requirements. Reaching tasks were ordered using a within-subjects experimental design, consisting of blocks of 30 reaches. A minimum of 30 seconds of rest was provided between blocks. Reaches were performed under each motion condition (Roll, Pitch, and Stationary). A white circular target (diameter = 1.27 cm) was presented in the center of one of three displays (Forward, Lateral, or Upward) in two visual conditions: Occluded Vision and Vision. The Occluded Vision task was a within-subjects experimental design consisting of six blocks, each containing 30 reaches. During each motion condition, the subjects experienced up to 180 seconds of continuous vibration exposure. After the participant returned their hand to the starting position, a target was presented. After visually acquiring the target, the participant simultaneously closed their eyes and depressed the home switch on the hand rest. After a discrete random delay (0, 500, 1000, 1500, or 2000 ms), an audible tone signaled the participant to begin the reach to the estimated target location. Reaction times were added to the delay times to create a continuous dataset from which to investigate the influence of effective visual occlusion on the fidelity of the mental representation of the target location. The Vision task also was a within-subjects experimental design consisting of six blocks, each containing 30 reaches under the same motion conditions.

114

DATA ANALYSIS Radial deviation of the fingertip was used to determine the off-axis direction of motion away from the intended trajectory at peak velocity (RDPV), and at the endpoint (RDEP). These directional errors were correlated as indicated below on a trial-by-trial basis, where low correlations would suggest evidence of feedback control (Feldman et al. 1994; Messier and Kalaska 1999). In the Occluded Vision condition, this feedback control would derive from non-visual, somatosensory feedback with respect to the memory-based internal model of the target location. To examine this, a radial mapping of the deviation of the fingertip away from the mean trajectory of replicated reaches was developed, where the deviations of the fingertip can be correlated between the peak velocity of the movement and the endpoint location of the fingertip on the display. A reach trajectory whose position at peak velocity is above that of the mean trajectory would be expected to end above the target location. If visual feedback mechanisms are not being utilized, than the dispersion of fingertip positions at the peak velocity (Figure 6-3b) should be replicated at the target. Conversely if proprioception is able to compensate for radial deviations, then endpoint locations at the target would be poorly correlated to the deviations at the peak velocity, as in Figure 6-3c.

115

Figure 6-3. a) Illustration of radial mapping used to identify deviations of fingertip position relative to the mean trajectory at the peak velocity (b) and the target (c).

Figure 6-4 further illustrates this concept, where reach endpoints are poorly correlated with their deviation at peak velocity. The mean trajectory is shown as the solid line connecting the reach origin to the display.

Peak Velocity Reach Origin Display

Figure 6-4. Illustration of the mapping of the deviation of the fingertip position at peak velocity to the reach endpoint on the display.

Two timing measures were recorded. Reaction time was recorded as the temporal delay between the stimulus and the release of the switch on the hand rest. Movement time was the time between the release of that switch and contact with the display. The reaction and movement times are aggregately referred to as the response time. A given trial was rejected if the recorded reaction time was less than 100 ms, or if the response time was outside the range of the participant

116

mean value plus five times the participant’s standard deviation, suggesting that the subject was not paying sufficient attention to that particular task.

Results Despite a practice session, a learning effect was observed as subsequent reaches yielded improved accuracy. When an order effect of increasing accuracy was found in the data, the first reach was removed from the dataset. This procedure was repeated until an order effect in response time was no longer significant. Hence approximately 25 reaches (out of 180) were removed from each participant’s dataset. Replicated trajectories were averaged to determine a “mean” trajectory of the intended movement (Faraway 2000). Deviations from this mean trajectory can be viewed as errors in movement planning or movement execution, but most often the latter (Soechting and Lacquaniti 1981). RIDE MOTION EFFECTS The analysis of variance with repeated measures was used to assess significance at the p < 0.05 level of each hypothesized main effect. The Ride Motion condition was shown to significantly affect reaction times, movement times, and endpoint accuracy of reaches performed in both the Vision and Occluded Vision conditions. The effects of ride motion on the endpoint variability of visually-occluded reaches are shown for each display in Figure 6-5, compared to the stationary situation. Ellipses depicted on the plots contain 95% of the endpoint dispersion. The area of the ellipses was two times greater in the ride motion condition than in the stationary condition. The principal component of the endpoint dispersion was consistent with the direction of the fingertip motion. For

117

example, Figure 6-5c shows that the endpoint dispersions for reaches to the Lateral display have an elongated orientation, because the hand is still moving left to right as the fingertip makes contact with the display, while there is little vertical movement of the hand (Gordon et al. 1995). Stationary

Ride Motion

(Units are millimeters)

a)

b)

c)

Figure 6-5. Endpoints without visual feedback under stationary and ride motion conditions to each of the three touchscreens: a) Forward-Up, b) Forward, and c) Lateral. Dashed ellipses contain 95% of the reach endpoints for that condition.

VISUAL OCCLUSION Peak tangential velocities (PTV) of the fingertip were significantly higher during visually-occluded reaches, corresponding to shorter movement times (Table 6-8). Ride motion contributed to a 6.3% increase in PTV in the Vision

118

condition, and an 11.5% increase in the Occluded Vision condition. Compared to reaches performed in the Stationary condition, Ride motion also resulted in a 25% increase in endpoint variability in the Vision condition, and a 32% increase in the Occluded Vision condition. Compared to the Vision condition, visual occlusion resulted in endpoint variability that was 2.3 times greater in the Stationary condition, and 2.5 times greater under ride motion. Table 6-8. Summary of dependent measures with respect to ride motion and visual feedback.

MT (ms) PTV (m/s) RDEP (mm)

Stationary 424 (85) 2.54 (0.9) 8.4 (6.5)

Vision Ride Motion 438 (119) 2.70 (0.9) 10.5 (7.2)

Occluded Vision Stationary Ride Motion 379 (85) 325 (73) 2.59 (0.8) 3.01 (0.9) 27.5 (16.2) 36.2 (19.2)

ANALYSIS OF FINGERTIP TRAJECTORY Using the circular representation shown in Figure 6-3, the endpoint variability of occluded reaches is shown for each display and motion condition in Figure 6-6. The radial deviation at PTV was hypothesized to be positively correlated with the radial deviation at the target with respect to the mean trajectory. However, correlations of these deviations were nearly zero: R2 values of -0.07, -0.03, and 0.06 for the Forward, Lateral, and Upward touchscreens respectively. Overall, the endpoint dispersion at the target are nearly zero correlation to values at the PTV for visually-occluded reaches (R2 = 0.07).

119

Lateral

Upward

Ride Motion

Stationary

Forward

Figure 6-6. Mapping of spatial dispersion of reach endpoints for each display (columns) and motion condition (rows).

DELAYS TO MOVEMENT ONSET Delay times between target stimulus and movement onset were analyzed separately as nominal and continuous variables. In the discrete 500-ms intervals (0 to 2000 ms), there was not a significant effect on the endpoint accuracy (p > 0.05), suggesting that memory-stored target information is not sufficient for performing an accurate reaching task. However, a significant effect was discovered when the time delay was modeled as a continuous variable including reaction time (Table 6-9). By including the reaction time as part of the delay to movement onset, a more precise estimate is obtained of the time required for the visuomotor system to remember the target location.

120

Table 6-9. ANOVA results for radial deviation of the fingertip at the target.

Source

N

DF

S.of S.

F Ratio

Prob > F

Subject (S)

9

9

8,769

3.36

0.0004

Ride Motion (RM)

2

2

5,416

9.34

< 0.0001

Display Location (DL)

2

2

1,689

2.91

0.055

Delay Time (DT)

1

1

2,166

7.47

0.006

S * RM

18

18

6,716

1.29

0.186

S * DL

18

18

20,630

3.95

< 0.0001

S * DT

9

9

4,229

1.62

0.104

RM * DL

4

4

1,907

1.64

0.161

RM * DT

2

2

837

1.44

0.236

DL * DT

2

2

265

0.46

0.634

Discussion This study shows that an unpredictable motion environment will degrade reaching movements through the biodynamic feedthrough of the vehicle motion to the fingertip. These findings are consistent with theories of movement planning and control that suggest that the ability to predict the motion environment is essential for the control of a reaching motion (Wolpert et al. 1995; Wolpert et al. 2000; Erlhagen et al. 2002). More specifically, to perform rapid pointing movements, the CNS must determine an initial reference frame based on available proprioceptive and visual information. Under random whole body motion, this reference frame may lose its fidelity due to sensory-mismatch and a decreased ability to predict the state of the environment and the target location. Hence, movement plans may be less accurate and require additional corrections to complete a reach. During visuallyguided reaching, it is believed that the CNS transitions its reference from the egocentric proprioception-based reference frame to an exocentric vision-based

121

reference frame. When vision is occluded, remaining somatosensory feedback mechanisms are used to compensate for large fingertip deviations, as evidenced by the radial deviation mapping shown earlier in Figure 6-6 and the near-zero correlations between deviations at the peak velocity and at the target. Since endpoint variability is much greater during occluded reaches, it is evident that non-visual feedback provides some, albeit insufficient, information to complete the reaching task. The near-zero correlation between radial deviation errors at the peak velocity and the reach endpoint shown in Figure 6-6 suggests that visuallyoccluded reaches are not executed in a strictly open-loop manner. Inference as to whether the feedbacks utilize memory-based or proprioception-based information is made primarily based on the significance of the movement onset delay, including reaction time, on endpoint accuracy. Although this study found no clear evidence of a latent target representation after brief delay intervals, as Westwood et al. (2003) suggests, it is suggested that proprioception is most likely the principal mechanism used to discern and modify hand and/or joint positions in visually-occluded reaching tasks. What follows is discussion regarding specific results. SIMULATED VEHICLE MOTION Timing results in this study are consistent with increased variability in the movement planning process. The motion environments introduce variability not only in the final position of the targets, but also in the orientation of the participants relative to gravity. This configuration requires a highly adaptive and

122

dynamic movement planning process. Such processes have been demonstrated to define movement plans along a continuum, in which the probability of an executed open-loop movement successfully reaching the goal can be related to the information available, and the resulting resolution of the movement plan (Bastian et al. 1998; McDowell et al. 2002). In the tasks examined here, ride motion was associated with increased reaction times; findings that are consistent with broader, more variable movement plans for these conditions. Participants exhibited longer movement times in the Vision condition then in the Occluded Vision condition, presumably because visual feedback provides additional information that enables increased accuracy but requires additional time for corrective submovements (Meyer et al. 1988). Ride motion resulted in an approximately 30% increase in endpoint variability compared to the Stationary condition. This consistent effect between visual conditions is not surprising as vibration is known to affect the precision of proprioceptive feedback, partially explaining the increased variability (Gauthier et al. 1981). Peak movement velocities were also higher under ride motion, possibly due to pre-movement muscular excitation, resulting in increased recruitment of motor units (Martin and Park 1997, Griffin et al. 2001). Interesting to note is that these visually-guided reaches with higher peak velocities under ride motion were also longer in duration, possibly due to increased muscular cocontraction that may be necessary to help stabilize the upper extremity during the reach in an unstable environment (Gribble et al. 2003; Milner 2002). In this theory, antagonist

123

muscles would be used for stabilization, opposing the intended movement and slowing the speed of the fingertip during the latter phase of the motions. END-EFFECTOR TRAJECTORY All of the successful rapid reaches contacted the touch-screens with a measurable velocity, as determined by kinematic VICON data. More specifically, the touch-screens abruptly stopped the distal movement of the arm, as opposed to commonly reported undershoot observed in unconstrained visually-occluded reaches (Flanders et al. 1992; Chieffi et al. 1999; Saunders and Knill 2004). Westwood et al. (2003) suggests that subjects may undershoot a target due to a poor representation of the target’s distance, or that they use a specific movement strategy to reduce effort or temporal demands. The subjects in this study may have overshot the target locations for similar reasons: a poor representation of the target after delays in the movement onset, increasing their speed and planned movement distance to ensure successful contact, without time-costly trajectory corrections. In other words, the results would indicate that the cost of undershooting was greater than that of overshooting, since the former would require additional movement corrections to reach the target, and hence lengthen movement time. In this study penalties were not assessed for missing the target on the touchscreen. Another possibility is that the CNS is simply attempting to solve the dual constraint of the speed/accuracy tradeoff of the task; simultaneously maximizing speed and accuracy (see Schmidt and Lee 1999 for review). In the Vision task, the CNS is able to use visual feedback to make trajectory modifications and

124

successfully complete the reaching task. In the Occluded Vision task, the absence of visual feedback inhibits the CNS’ ability to evaluate the trajectory and assess the need for online corrections. Without vision’s contribution to determining reach accuracy, the CNS places emphasis on being fast, where accuracy is poorly discerned. Additionally the small correlation between endpoint variability and velocity at target contact agree with several studies showing that direction and extent of reaching movements can be planned separately (Soechting and Flanders 1989; Favilla and De Cecco 1996; Messier and Kalaska 1997). As was observed with the ride motion effect, visual occlusion also had a consistent effect on endpoint accuracy across participants, 3.3 times more variable in the Occluded Vision task. Peak tangential velocities are shown to vary with the availability of visual feedback. Current movement theory suggests that without the use of visual feedback, deviations in rapid, visually-occluded reach trajectories are not corrected due to poor resolution of remaining afferent feedbacks (Messier and Kalaska 1997; Heath 2005). Interestingly Feldman et al. (1994) adds that control processes end before peak tangential velocity, but this fundamentally excludes feedback control processes. If these hypothesis are true, then it follows that visually-occluded movements would exhibit endpoint variability that is highly correlated with initial directional errors. However the results presented here show that initial direction errors at peak velocity are not correlated (R2 = 0.07) to the radial deviation errors at the target, suggesting that the CNS is able to use proprioceptive information to recognize and compensate for large deviations. Contrary to other findings (Heath 2005, Westwood et al.

125

2003), evidence of proprioceptive-based movement corrections suggests that these

visually-occluded

movements

are

not

completely

planned

before

commencing the motion. The absence of vision can have significant impact, especially in temporally demanding situations, such as high-speed driving. These scenarios typically contain in-vehicle reaching tasks that are referred to as secondary tasks, where the primary task is lane-keeping or some other mission-critical objective. Vision must be maintained on the primary task as much as possible. When a reach (secondary task) must occur, the occupant visually acquires the target and begins the reach, but then often returns gaze to the primary task before completing the reach. The results of this study indicate that in these late-reach tasks, the absence of visual feedback can cause significant degradations in reach performance. The successful completion of these reaches then relies on proprioceptive feedback which has been shown to be insufficient. Additional sensory information is required for the occupant, such as haptic feedback to complete these reaches. Information on target shape, size and texture become necessary to complete the task. Designing controls and displays that provide additional sensory information, even in multi-modal capacities, could substantially increase task performance.

126

Chapter 7 Modeling active human biodynamic response during reaching This chapter integrates the experimental findings reported in the previous chapters. It details the manual control strategies for planning, executing, and controlling the fingertip trajectories used during reaches to push-button targets within a vehicle being subjected to motion perturbations. It describes how these strategies are modified by the passive biodynamic response of the seated operator to vehicle ride motions, and the possible active neuromuscular control used to compensate for trajectory deviations.

Background Understanding the capabilities and limitations of human movement in a dynamic environment is essential in order to effectively design controls and displays for use in moving vehicles. Controls and displays must be easily accessible, simple to manipulate, and provide immediate feedback regarding the success or failure of the intended task. For many years, physical prototypes (costly in both time and money) have been used to evaluate designs with respect to human performance. More recently, virtual prototypes and digital human modeling (DHM) have become plausible alternatives that enhance the ability to predict and simulate human motions and behaviors (Chaffin 2001). DHM applications have also assisted in the ability to graphically represent human movement with the manipulation of digital humans, such as the Jack (Unigraphics 127

2006) and SAFEWORK avatars (SAFEWORK 2006). Although visually adequate representations can be made with relative ease, the validity of postures and motions is essential to any movement analysis (ibid.). Much work has been done to simulate and predict accurate human motor behavior, particularly in the performance of seated and standing tasks (i.e. Hsiang and Ayoub 1994; Zhang and Chaffin 2000; Faraway 2000, 2001, 2003; Abdel-Malek et al. 2004). Unfortunately typical DHM applications involve stationary environments, drastically different from those often experienced in military, transportation, and construction industries for example. Workers in these fields are often exposed to varying levels of whole-body vibration (WBV). Some biodynamic, or mechanistic, models utilize transfer functions and inertial properties of the human to simulate these scenarios (Matsumoto and Griffin 2001). Predictions of hand trajectories (e.g. Faraway 2001) are extrapolations based on empirical data collected on reaches to nearby locations. While these predictions inherently contain the feedback utilized when the reaches were performed, the hand trajectories are effectively open-loop predictions, as the models are incapable of responding to environmental changes during the reach. Two additional components must be incorporated in the reach prediction in order to simulate reaches in dynamic environments: biodynamic response, and feedback control. As discussed in the first two chapters, and then supported by the empirical results presented in chapters 3 through 5, terrain-induced ride motion induces perturbations to movements that disrupt the intended reach trajectory. The passive biodynamic response of the seated vehicle operator is

128

predictable, at least in a biomechanical sense (Coermann 1962, Griffin 2001). The contributions of visual and proprioceptive feedbacks were discussed in Chapter 6. These feedbacks can be used to monitor perturbations and compensate for trajectory deviations (Schmidt and Lee 2005). If such compensatory mechanisms are included in reach prediction models, then through simulation of manual performance in ride motion environments, designers can efficiently design controls and displays to ensure that the operator’s movements are less susceptible to vehicle motion. Simulation of human movements is complex due to numerous interactive factors that cannot always be predicted. Some components are psychophysical in nature, such as motivation. Some have biomechanical underpinnings that include the infinite number of feasible solutions resulting from redundant degrees of freedom in normal human movements (Scholz and Schöner 1999). Still other factors cause natural variability, including neuromuscular noise (Schmidt et al. 1979; Ghez 1991) and unknown intersubject differences (Meyer et al. 1982). Incorporating these sources of variability where possible, hypotheses and models, such as the impulse-timing (Woodworth 1899; Bernstein 1967) and the equilibrium-point (E-P) hypothesis (Asatryan and Feldman 1965; Feldman 1966; Polit and Bizzi 1978, 1979), have been developed to explain and predict hand trajectories during reaching movements. In this chapter, an active human biodynamic response model is proposed, which generates a trajectory path of the right index fingertip during a rapid reach to an in-vehicle pushbutton target under ride motion perturbation. Elements of the

129

model are derived from motor control theory and human vibration responses documented earlier. An initial, open-loop trajectory is estimated based on the initial orientation of the body, including the position of the hands, and the location of the target. Movement variability is incorporated by modifying the departure angle of the predicted trajectory. Vibration feedthrough to the hand is simulated through an external biodynamic response model, and affects the entire reaching movement. Closed-loop feedback is modeled using a preview controller, but is dormant until the fingertip enters the field-of-view. These individual components are discussed below, as well as future plans to validate and improve the model. TRAJECTORY PLANNING AND EXECUTION In its simplest form, planning a reach requires several task parameters to be considered, including the following: starting hand position, desired end location, and any obstacles between the two locations. Several theories exist regarding

trajectory

formation,

though

much

of

the

literature

focuses

predominantly on Morasso’s seminal work (1981), which applies to “straight-line” movements with bell-shaped speed profiles in the absence of obstacles. These linear hand paths, however, may not be adequate representations of the actual three-dimensional hand trajectories as shown in earlier chapters, and by others (Atkeson and Hollerbach 1985; Breteler et al. 1998, Faraway 2001). Soecting and Lacquaniti (1981) have shown that pointing tasks have invariant kinematic characteristics, including the consistency in the trajectory of the hand regardless of movement speed. While replicated reaches are highly repeatable, Schmidt and Lee (2005) discuss the natural variability that prevents

130

repeated reaches from having identical kinetic or kinematic characteristics. They further suggest that this variability is introduced at two levels: movement planning and movement execution. Before a movement is initiated, an inexact estimation of the exocentric target location or egocentric postural representation errors create a discrepancy between a hypothetically optimal movement plan and the planned trajectory (Soechting and Flanders 1989a). Repetition and learning reduce but do not eliminate this effect. In addition, errors in movement execution result from the imperfect execution of the intended movement plan, due to variability in the neuromuscular system and biomechanical perturbations induced by a motion environment. BIODYNAMIC PERTURBATION AND MODEL In the latter regards, intended movements are often disturbed by external influences while extending the arm during a reach. Particularly in the case of moving vehicles, ride motions are transmitted through the vehicle subsystems to the operator, which causes a biodynamic response of the extended hand referred to as vibration feedthrough. These terrain-induced ride motions and biodynamic human responses are incorporated in three-dimensional vehicle and biodynamics models, which were discussed in Chapter 3. One example is AVB-DYN, developed by Systems Technology, Inc. (STI, Hawthorne, California), containing an independent biodynamic response model (BIODYN) that calculates the dynamic response of a seated operator in a given posture, when subjected to ride motion. The AVB-DYN package incorporates a tire model, vehicle suspension, and seating modules to determine the vibration accelerations that are transmitted

131

to the vehicle-occupant coupling, namely the seatback, seatpan, floor, and steering wheel. These accelerations are input to the BIODYN module, which calculates the human biodynamic response of the seated operator using a set of experimentally derived estimates for the kinetic parameters shown in Figure 7-1 (Allen et al. 2005). In this regard, the model can be used to evaluate the passive biodynamic response while the operator is gripping a steering wheel.

Figure 7-1. Example of a dynamic kinetic model of a seated vehicle operator used to calculate the passive biodynamic responses to sagittal plane ride motion while holding a steering wheel (from Allen et al 2005).

VISUAL AND SOMATOSENSORY FEEDBACKS Visual and somatosensory feedbacks are the two principal mechanisms contributing to the regulation of reach performance (Bernstein 1967; Jeannerod 1988; Riemann and Lephart 2002). Without visual feedback, reach accuracy is known to degrade (Keele and Posner 1968, Zelaznik et al. 1983). Likewise the absence of proprioception (Ghez et. al 1990; Gordon et al. 1995) and the

132

vibration-induced alteration of proprioception (Gauthier et al. 1981, Martin et al. 1991) reduce reach performance. The model described in the following sections incorporates continuous feedback that transitions from proprioceptive to visual feedback during each reach being simulated, based on the remaining distance between the fingertip and the target. Vision becomes the dominant feedback mechanism after the end-effector enters the field-of-view (Schneider 1969; Blouin et al. 1993), though it is estimated to take between 85 and 135 ms for the visuomotor system to execute a compensatory motor command based on visual feedback information (Soechting and Flanders 1989b; Carlton 1992). The normal stereoscopic visual field is approximately 200° horizontally, and 135° (+60°, -75°) vertically, although the central visual processing is closer to ± 60° horizontally, and only 75° (+30, -45°) vertically (Gibson 1979; Werner 1991; Paillard 1996). Assuming that peripheral vision can provide an estimation of the fingertip movement (Jagacinksi and Flach 2003), the proposed model utilizes visual feedback when the fingertip enters the central visual field. PREVIEW CONTROL AND TRAJECTORY MODELING MacAdam (1980), Prokop (2001), and Gordon et al. (2002) have developed vehicle-driver models that utilize a “look-ahead”, or preview distance that achieves increased stability in a dynamic system by modifying the system state based on anticipated events, effectively removing the time-delayed response. These optimal control models utilize a fixed preview distance, which provides the desired position fingertip position at some future time. The preview controller evaluates the differences between the current position and velocity, and the look-ahead vector. It

133

then generates accelerations that will satisfy the desired position. Jagacinski and Flach (2003) discuss a similar notion of “quickening”, or the ability to reconcile current position and velocity information to predict future positions. While B-splines have commonly been used to simulate the changes in movement trajectories between known points (Faraway 2001), the trajectory controller proposed in this algorithm utilizes cubic splines in Hermite polynomial form. B-splines use control points that define the curvature of the movement path, and are quite similar to Hermite curves in this regard. However B-spline control points are typically arbitrarily defined, whereas the “control points” of Hermite splines are vectors tangential to the current and desired directions. These tangential vectors relate directly to the current velocity vector, and the desired vector orthogonal to the target plane; both objectively defined and more physically relevant to human movement (Mori and Hoshino 2002). The proposed trajectory control algorithm simulates the hand trajectory of target-directed reaching tasks, perturbed by ride motion, incorporates natural movement variability, and includes a feedback controller that compensates for vibration feedthrough resulting from vehicle ride motion. The following subsections outline the structure of the algorithm, and how its components integrate into the current model development.

Simulating Trajectories Observed under Ride Motion (STORM) A general schematic of the STORM concept is presented in Figure 7-2, while the algorithm architecture is presented in Figure 7-8, following individual descriptions of the algorithm components.

134

135

Figure 7-2. Schematic of proposed feedback control model (STORM) for predicting fingertip trajectories during reaching motions in a vehicle subjected to ride motion perturbations.

In order to simulate reach trajectories in a dynamic environment, STORM requires that several inputs be known a priori. At a minimum, these inputs include the subject’s stature and weight, an initial starting posture, the target location, the ride motion profile at the seat, and the spatiotemporal demands of the task given by a number inclusively between 0 (accuracy) and 1 (speed). This value, referred to

as

“S-A”

representing

the

speed-accuracy

tradeoff,

modulates

the

performance observed under various spatiotemporal constraints. Secondly, a vibration dose value (VDV) is calculated from the vibration inputs to quantify the direction, frequency, and magnitude of the ride motion accelerations into a single numerical value. The spatiotemporal weighting and the VDV value are used in the modified formulation of Fitts’ law, presented in Chapter 5 to determine the expected movement time for the reach under these dynamic conditions and spatiotemporal constraints. Based on the results presented in Chapter 5, the endpoint accuracy of the simulated reaches range from precise contact (S-A = 0) to an open-loop movement (S-A = 1). To begin the simulation, UGS Jack is used to position the seated operator in a nominal posture of a person driving a High Mobility Multipurpose Wheeled Vehicle. A Jack module was specifically developed to export anthropomorphic and postural data into the BIODYN model for biodynamic calculations. BIODYN also incorporates data from GEBOD that generates geometric and body segment masses, and the locations and mechanical properties of joints for humans (Cheng et al. 1994). Based on this initial posture and the target location, a sequence of whole-body seated postures is predicted emulating the intended

136

movement to the target location (Faraway 2001). The prediction is an interpolation of seated reaches performed in a stationary environment, and generally includes any reasonable locations within the functional reach envelope. The trajectory predicted for a reach to a specific target is then modified using a normal distribution of radial deviations at peak velocity to simulate human movement variability, and result in a prediction of the executed open-loop movement. This movement is valid until it is perturbed by vibration feedthrough, or modified via feedback control. These deviations at the peak velocity form departure vectors, or launch cones, from the origin towards the target (Figure 7-3).

Origin Target

Figure 7-3. Illustration of normally distributed “launch cone” incorporating the initial trajectory for 95% of empirical reaches to a target location.

The launch cones are a planar distribution of the fingertip location at the peak velocity, orthogonal to reach direction of the predicted trajectory. Distributions are illustrated in Figure 7-4 as dashed ellipses, generally circular, and contain approximately 95% of the data points under these conditions. Areas of the ellipses are provided on the plots, and were greater for reaches to each display under ride motion, when compared to reaches performed in the stationary cab.

137

A = 0.35 m2

A = 0.63 m2

A = 0.46 m2

A = 1.14 m2

A = 1.23 m2

A = 1.12 m2

Figure 7-4. Comparison of spatial dispersions of the fingertip at the peak velocity in a stationary cab (Top), and under ride motion (Bottom). Plots are for reaches to the Forward, Lateral, and Upward displays (respectively left to right).

Atkeson and Hollerbach (1985) and Gordon et al. (1994) suggest that executed movements are initially open-loop and are not corrected until an error is detected – sometime after the hand attains peak velocity (Messier and Kalaska 1999). The STORM algorithm uses two random numbers to simulate the stochastic variation of movement execution. The first number is randomly selected from a uniform distribution between 0 and π, which determines the radial axis of deviation from the reference trajectory. The second number is selected from a normal distribution (mean and standard deviation based on the ellipses shown in Figure 7-4), and equates to the magnitude of the deviation away from the center. The deviation calculated from these numbers is then used to adjust the entire reference trajectory using a linear extrapolation (Figure 7-5).

138

This adjustment provides an estimate of the effect of neuromuscular variability on the executed open-loop trajectory. Reference trajectory Gradient-adjusted trajectory

Figure 7-5. Adjustment to reference trajectory based on linear extrapolation of the deviation at the peak velocity.

In the case of reaching in a dynamic motion environment, the vibration feedthrough perturbs the open-loop trajectory immediately after the hand releases the steering wheel. BIODYN imports the anthropomorphic and postural data from Jack and calculates the instantaneous perturbation to the hand, based on the vibration feedthrough of the vibration input at the seatpan, backrest, floor, and steering wheel to the left hand. This perturbation is given as an acceleration vector of the hand, and is then summed with the instantaneous acceleration of the volitional hand movement to determine the subsequent hand location at the next point in time. As movement times have been shown in previous chapters to vary as a function of the ride motion perturbation, this new hand location will be different from the initially predicted location at this time in the reach trajectory. The original prediction will likely be a close approximation, but will be temporally inaccurate. Therefore, the torso and arm linkage is modified slightly in the Jack environment to produce a more representative posture of the current trajectory position using the new hand location and H-point of the seated avatar as endpoints for Faraway’s stretch-pivot method (2003). This new posture is then exported to BIODYN, where the process is repeated 30 times per second, per Jack’s simulation clock, until the 139

reach is successfully completed. An example illustrating the superposition of hand accelerations over an entire trajectory is shown in Figure 7-6. Superposition of manual and passive accelerations in vertical axis

Acceleration (m/s^2)

Volitional control

Feedthrough

Observed

25 0 -25 0

100

200

300

on (m/s^2)

Superposition of manual and passive accelerations in lateral axis

Acceleration (m/s^2)

Acceleration (m/sec^2) Accelerati

Time (ms)

25 0 -25 0

100

200

300

Time (ms) Superposition of manual and passive accelerations in longitudinal axis 25 0 -25 0

100

200

300

Time (ms)

Figure 7-6. Typical example of superposing empirical fingertip trajectories and vibration feedthrough accelerations (Rider and Martin 2005).

Finally, the resulting fingertip trajectory vector is monitored by a feedback controller that calculates the future position of the fingertip to determine if the continuation of the current trajectory will successfully hit the target without 140

modifications. The preview control is modeled here as a Hermite curve interpolation between the current movement vector and the landing vector, orthogonal to the target plane (Rider and Martin 2005). The tangential vectors generate a kinetic smoothness similar to natural movement, as illustrated in Figure 7-7. The Hermite form consists of two control points and two control tangents for each polynomial, where the trajectory is subsequently defined by the transition between current and desired states. “On each subinterval, given a starting point (p0) and an ending point (p1) with starting tangent (m0) and ending tangent (m1), the polynomial can be defined by the following equation” (Wikipedia 2006):

p(t) = ⎛⎜ 2t 3 − 3t 2 + 1 ⎞⎟ p + ⎛⎜ t 3 − 2t 2 + t ⎞⎟m + ⎛⎜ − 2t 3 + 3t 2 ⎞⎟ p + ⎛⎜ t 3 − t 2 ⎞⎟m ⎝ ⎠ 0 ⎝ ⎠ 0 ⎝ ⎠ 1 ⎝ ⎠ 1 The four Hermite basis functions can be defined by the following equations: h (t ) = 2t 3 − 3t 2 + 1 00 h (t ) = −2t 3 + 3t 2 01 h (t ) = t 3 − 2t 2 + t 10 h (t ) = t 3 − t 2 11

Numerically, these equations can be parsed in matrix form, where the projected trajectory is the multiplication of terms (Pipenbrinck 2006): P = S * h * C, where

⎡s 3 ⎤ ⎢ 2⎥ s S=⎢ 1⎥ ⎢s ⎥ ⎢ ⎥ ⎣⎢ 1 ⎦⎥

1⎤ ⎡ 2 −2 1 ⎢− 3 3 − 2 − 1⎥ ⎥ h=⎢ ⎢0 0 1 0⎥ ⎢ ⎥ 0 0 0⎦ ⎣1

⎡ P0 ⎤ ⎢P ⎥ C = ⎢ 1 ⎥. ⎢M 0 ⎥ ⎢ ⎥ ⎣ M1 ⎦

141

Control points Tangential vectors Hermite spline

Origin

Target

Figure 7-7. Illustration of the Hermite curve using the tangential vectors of current and desired velocity vectors.

The STORM algorithm incorporates the iterative superposition of the executed trajectory, vibration feedthrough, and feedback control to produce a fingertip trajectory that will satisfy the spatial requirements of the target. While the algorithm is continually monitoring the relative position between the fingertip and the target, the Hermite feedback controller only operates when the projected fingertip endpoint is outside of the effective target width and the fingertip is located within the central visual field, as discussed earlier in this chapter. The architecture of the STORM algorithm is illustrated in Figure 7-8, representing the data flow through each component of the STORM framework.

142

Position digital human in nominal driving posture

Start

Calculate expected MT from modified Fitts’ law

Calculate VDV and effective target width ETW from S-A

Minimal inputs: 1. Stature 2. Weight 3. Initial posture 4. Target location 5. Motion profile 6. Spatiotemporal constraint (S-A)

Determine launch cone and adjust target location Model outputs: 1. Reach trajectory 2. Movement time 3. Expected accuracy

Generate sequence of postures for trajectory (Faraway 2001) Export anthropomorphic data to BIODYN

Calculate acceleration of hand resulting from biodynamic feedthrough End

Superimpose hand accelerations of volitional and biodynamic response

yes Is the fingertip at the target?

Adjust predicted posture using stretch-pivot method (Faraway 2003)

Is fingertip within central field of view?

no

Calculate modified fingertip trajectory based on Hermite spline

no yes

yes Calculate reach endpoint based on current path

Will current trajectory hit ETW?

no

Figure 7-8. Architecture of STORM algorithm, and component integration.

143

Chapter 8 Summary and future research opportunities Ride motion has been shown to significantly degrade the speed and accuracy of in-vehicle reaching and pointing tasks. The effects of ride motion and visual occlusion on the spatial dispersion of reach endpoints presented in Chapters 5 and 6. Ride motion produced a 30% increase in endpoint variability in temporally-unconstrained tasks, and a 56% increase for rapid reaching tasks. In the stationary condition, the absence of visual feedback contributed to endpoint variability that was approximately 3.3 times greater. Collectively, ride motion perturbations to visually-occluded reaches were almost 4.5 times greater than visually-guided reaches in a stationary cab. The STORM algorithm provides mechanisms by which to simulate openand closed-loop reach trajectories, and is structured to simulate reach trajectories in a stationary or dynamic environment. Simulated reaches using only the trajectory prediction, neuromuscular variability, and feedback control components resemble the visually-guided, stationary cab reaches. Inclusion of the vibration feedthrough enables simulating visually-guided reaches performed under ride motion. Furthermore, removal of the feedback control model would simulate the open-loop, visually-occluded movements described in Chapter 6. As future research provides a better understanding and improved models of movement planning and control, including active and passive human

144

biodynamic responses, the relevant components in the STORM algorithm can be updated or replaced. Regardless of each component, it is suggested that one of the most significant contributions is the architecture of the algorithm. Its structure promotes system-level analyses of human movements in dynamic environments. For example, BIODYN currently uses the same joint stiffness parameters throughout a reach, as it was originally designed to calculate the biodynamic response of a maintained posture, despite McRuer’s et al. (1965) earlier work suggest that humans adjust the stiffness of their joints to create a consistent transfer function from the vibration input to the end effector in an attempt to improve their manual coordination. Though BIODYN utilizes joint stiffness parameter estimates from empirical data, such as the work presented in Chapters 3 and 4, further research on variable joint stiffness parameters is needed and could be included in the STORM algorithm. The evaluation of many human-centered design scenarios must include an assessment of the speed and accuracy tradeoff, which is almost entirely taskdependent. In cases where space is available and temporal constraints are relaxed, general recommendations for improving the performance of a reaching task include the following: 1. Minimize vibration transmission to vehicle occupants to reduce the degradation of manual performance tasks, 2. Target size should relate to reach distance, accounting for the direct relationship between reach distance and endpoint variability,

145

3. Increased separation between critical controls, to increase the likelihood of a successful task, and reduce likelihood of costly errors. 4. Place target in locations where reaching movements are less susceptible to motion perturbations, and thus increasing accuracy, and 5. Orient target so that it can be easily seen, with an unobstructed path, so that it can be more quickly and easily reached. In cases where speed is essential, recommendations 1 and 3 remain, recognizing that larger targets are typically not an acceptable solution in spaceconstrained environments, such as the cockpits of vehicles. An arm or wrist rest may provide additional stability to press targets more easily, although it may likely result in increased movement times for single target-aimed reaching tasks. In the case of sequential inputs at the same location, such as keypad data entry, wrist rests provide increased stability of the extended arm and hand, attenuating the magnitude of vibration feedthrough to the hand, and improve task performance (i.e. shorter task times). During the use of touchpanel displays, a design tradeoff should be evaluated based on the amount of information that needs to be provided, and any task-related time constraints. Immediate and informative feedback should be given when a displayed target is successfully pressed.

Summary of Findings Whole-body vibration (WBV) adversely affects the speed and accuracy of in-vehicle reaching tasks. Under moderate vibration magnitudes, this translates to 13.5% longer reaction times to stimulus presentation, 10% longer movement times in temporally-unconstrained movements, and a 56% decrease in endpoint

146

accuracy. Specifically, the resonance effects of the torso to low frequency vibrations correlate with increased variability of the extended hand; this variability, or effective target width (ETW), is highly dependent on the location of the control or display. ETW is 11.1% greater when the principal direction of WBV is perpendicular to the reach direction, as compared to 6.6% increases when the reach direction coincides with the vibration direction. Furthermore, reaches to locations that require use of the torso result in 11.6% longer movement times than proximally located targets. The ETW also increased proportionately as reach distance increased – approximately 8% of the reach distance in unobstructed reaching tasks. Reaches to physical and digital pushbutton targets exhibited success rates that were not significantly different; however reaches to 12.7 cm (0.5”) targets were almost 14.8% faster to physical buttons, while reaches to 25.4 cm (1.0”) were faster to digital targets by 6.3%. On the other hand, “miss” rates for rapid reaches were approximately 3 times greater under moderate ride motion than rapid reaches performed in a stationary environment. Movement times under rough ride motion were 60% greater than reaches performed under moderate levels of vibration, reaches which in turn were 60% longer than those performed under low vibration levels. Lastly, these empirical results have been integrated into an algorithm, through which a systems-level analysis can be performed to assess the expected reach performance of in-vehicle tasks under of ride motion perturbation. This algorithm includes movement planning and variability, passive and active biodynamic feedthrough, and neuromuscular feedback control processes to successfully complete the intended task.

147

There are of course additional factors contributing to the variability in movement time and accuracy, such as anthropometry, coordination, and motivation, which must be investigated further. The within-subject variation must be more thoroughly addressed and measured to provide the needed predictive power to develop a robust dynamic model of reach capability. Ideal design recommendations may have to include a priori knowledge of the anthropometry and coordination of the user. Generating design guidelines remains a challenge; however gaining sufficient understanding for the prediction and simulation of human movement in dynamic environments is increasingly plausible. The ability to model the passive and active biodynamic human responses in a dynamic environment provides a method to investigate how people will perform novel manual tasks in unpredictable and unstable environments. These human models can be integrated into existing simulations of current and proposed environments, where manual performance can be evaluated before placing a person in a potentially harmful situation. Predicting operator performance then becomes an invaluable component in the design of future passenger, commercial, and military vehicles, where in-vehicle tasks must be completed despite the dynamic coupling of the vehicle to unpredictable terrain. As these biodynamic human models improve in fidelity, the capability and limitations of human movement will become more transparent and predictable. The knowledgebase that results will enable improved human-centered designs of vehicle controls and displays, which will minimize the deleterious effects of the ride motion environment on in-vehicle reaching tasks.

148

References Abdel-Malek K, Yu W, Jaber M, Duncan J (2001) “Realistic Posture Prediction for Maximum Dexterity,” SAE Digital Human Modeling Conference, Arlington, VA, 2001-01-2110. Abend WK, Bizzi E, Morasso P (1982) “Human arm trajectory formation,” Brain 105: 331-348. Adamovich S, Archambault P, Ghafouri M, Levin M, Polzner H, Feldman A (2001) “Hand Trajectory invariance in reaching movements involving the trunk,” Experimental Brain Research 138: 288-303. Adamovich SV, Berkinblit MB, Fookson O, Poizner H (1998) “Pointing in 3D space to remembered targets. I. Kinesthetic versus visual target presentation,” Journal of Neurophysiology 79(6): 2833-2846. Adamovich SV, Levin MF, Feldman AG (1994) Merging different motor patterns: coordination between rhythmical and discrete single-joint movements. Exp Brain Res 99:325–337 Adamovich SV, Burlachkova NI, Feldman AG (1984) Wave nature of the central process of formation of the trajectories of change in the joint angle in man. Biophysics 29:130–134 Allen RW, Klyde DH, Thompson PM, Magdaleno RE, Liang C-Y, Rosenthal TJ (2005) “An Integrated Anthropometrics, Vehicle, and Biodynamics Software Tool,” STI-TR-1354-1, Systems Technology, Inc., Hawthorne, CA, 13 May 2005. Allen RW, Jex HR, Magdaleno RE (1973) “Manual control performance and dynamic response during sinusoidal vibration,” AMRL-TR-73-78, October. Asatryan DG, Feldman AG (1965) “Biophysics of complex systems and mathematical models. Functional tuning of nervous system with control of movement or maintenance of a steady posture – I. Mechanographic analysis of the work of the joint on execution of a postural task,” Biophysics 10: 925-935. Atkeson CG, Hollerbach JM (1985) “Kinematic features of unrestrained vertical arm movements,” J Neuroscience 5(9): 2318-2330. Baird KM, Hoffman ER, Drury CG (2002) “The effects of probe length on Fitts’ law,” Applied Ergonomics 33(1): 9-14.

149

Bastian A, Riehle A, Erlhagen W, Schöner G (1998) Prior information preshapes the population representation of movement direction in motor cortex. NeuroReport 9: 315-319. Bédard P, Proteau L (2001) “On the role of static and dynamic visual afferent information in goal-directed aiming movements,” Experimental Brain Research 138: 419-431. Beggs, W. D. A.; Howarth, C. I. (1972) “The movement of the hand towards a target,” Quarterly Journal of Experimental Psychology, 24: 448-453. Benel, R. A.; Stanton, B. C. (1987) “Optimal size and spacing of touch screen input areas,” Human-Computer Interaction - INTERACT, 581-585. Berkenblit M, Fookson O, Smetanin B, Adamovich S, Poizner H (1995) The interaction of visual and proprioceptive inputs in pointing to actual and remembered targets. Exp Brain Res 107: 326-330 Bernstein N (1967) In: The coordination and regulation of movements. Oxford, New York: Pergamon Press. Bittner A, Guignard J (1985) “Human factors engineering principles for minimizing adverse ship motion effects: Theory and practice,” Naval Engineers Journal 97: 205-213. Blouin J, Teasdale N, Bard C, Fleury M (1993) Directional control of rapid arm movements: the role of the kinetic visual feedback system. Can J Psych 47: 678696. Boileau, P.; Rakheja, S. (1998) “Whole-body vertical biodynamic response characteristics of the seated vehicle driver: Measurement and model development” International Journal of Industrial Ergonomics, 22: 449-472. Boileau, P.; Rakheja, S.; Yang, X.; Stiharu, I. (1997) “Comparison of biodynamic response characteristics of various human body models as applied to seated vehicle drivers” Noise and Vibration Worldwide, 28(9): 7-15. Bovenzi M, Hulshof CTJ (1998) “An updated review of epidemiologic studies on the relationships between exposure to whole-body vibration and low back pain,” Journal of Sound and Vibration 215(4): 595-611. Brenner, E., Smeets, J. (1995) “Moving one’s finger to a visually specified position: target orientation influences the finger’s path,” Exp. Brain Res. 105: 318-320. Bresciani JP, Blouin J, Popov K, Bourdin C, Sarlegna F, Vercher JL, Gauthier GM (2002) “Galvanic vestibular stimulation in humans produces online arm

150

movement deviations when reaching towards memorized visual targets, Neuroscience Letters 318: 34-38. Breteler, M.; Gielen, S.; Meulenbroek, R. (2001) “End-point constraints in aiming movements: Effects of approach angle and speed,” Biological Cybernetics, 85: 65-75. Buck L (1982) “Location versus distance in determining movement accuracy,” Journal of Motor Behavior 14: 287-300. Carlton LG (1992) Visual processing time and the control of movement. In: Proteau L & Elliott D (eds) Vision and motor control. Amsterdam: North-Holland. Carlton LG (1981) “Processing visual feedback information for movement control,” Journal of Experimental Psychology: Human Perception and Performance 7: 1019-1030. Carson RG, Goodman D, Chua R, Elliott D (1993) Asymmetries in the regulation of visually guided aiming. J Mot Behav 25: 21-32 Cesari, P.; Shiratori, T.; Olivato, P.; Duarte, M. (2001) “Analysis of kinematically redundant reaching movements using the equilibrium-point hypothesis,” Biological Cybernetics, 84: 217-226. Chaffin DB (2001) In: Digital human modeling for vehicle and workplace design. Society of Automotive Engineers, Warrendale, PA. Cheng H, Obergefell L, Rizer A (1994) In: Generator of Body Data (GEBOD) Manual, AL/CF-TR-1994-0051. Clower DM, Hoffman JM, Votaw JR, Faber TL, Woods RP, Alexander GE (1996) “Role of posterior parietal cortex in the recalibration of visually-guided reaching,” Nature 383(6601): 618-621. Coermann RR (1962) “The mechanical impedance of the human body in sitting and standing positions at low frequencies,” Human Factors 4: 227-253. Cohen, H.H.; Wasserman, D.; Hornung, R. (1977) "Human Performance and Transmissibility under Sinusoidal and Mixed Vertical Vibration," Ergonomics, 20(3): 207-216. Contreras-Vidal J, Kerick S (2004) “Independent component analysis of dynamic brain responses during visuomotor adaptation,” NeuroImage 21: 936-945. Corbridge C, Griffin MJ (1991) “Effects of vertical vibration on passenger activities: Writing and drinking,” Ergonomics 34(10): 1313-1332.

151

Corbridge C, Griffin MJ (1986) “Vibration and comfort: Vertical and lateral motion in the range 0.5 to 5.0 Hz,” Ergonomics, 29(2): 249-272. Cowings P, Toscano W, DeRoshia C, Tauson R (1999) “The effects of Command and Control Vehicle (C2V) operational environment on soldier health and performance,” (Rep No ARL-MR-468) Aberdeen Proving Grounds, MD: Army Research Laboratory. Crossman EREW, Goodeve EJ (1983) “Feedback control of hand-movement and Fitts' law,” Quarterly Journal of Experimental Psychology, 35A: 251-278. Desmurget M, Pélisson D, Rossetti Y, Prablanc C (1998) From eye to hand: Planning goal-directed movements. Neurosci Behavioral Review 22: 761-788 Dewhurst DJ (1967) “Neuromuscular control system,” IEEE Transactions on Biomedical Engineering 14: 167-171. Elliott D, Binsted G, Heath M (1999) “The control of goal-directed limb movements: Correcting errors in the trajectory,” Human Movement Science 18: 121-136. Elliott D, Carson RG, Goodman D, Chua R (1991) “Discrete vs. continuous visual control of manual aiming,” Human Movement Science 10: 393-418. Elliott D, Madalena J (1987) The influence of premovement visual information on manual aiming. Quarterly J Exp Psych 39A: 541-559 Erlhagen W, Schöner G (2002) “Dynamic field theory of movement preparation,” Psychol Rev 109: 545-572. Fairley TE, Griffin MJ (1989) “The apparent mass of the seated human body: vertical vibration,” J Biomechanics 22(2): 81-94. Faraway JJ (2004) In: Linear Models with R. Chapman & Hall/CRC Press. Faraway JJ (2003) “Data-based Motion Prediction,” Journal of Passenger Cars Electronic and Electrical Systems: 722-732. Faraway JJ, Hu J (2001) “Modeling Variability in Reaching Motions,” Technical Paper 2001-01-2094. Warrendale, PA: SAE International. Faraway JJ (2000) “Modeling reach motions using functional regression analysis,” Technical Paper 2000-01-2175. Warrendale, PA: SAE International. Feldman A, Levin M (1995) “The origin and use of positional frames of reference in motor control,” Behav Brain Sci 18: 723-806

152

Feldman A, Adamovich S, St-Onge N, Levin M (1994) Control processes in fast movements end before the peak velocity: evidence based on experimental analysis and modeling. Proceedings of the 16th annual conference of the IEEE EMBS, Baltimore, MD Feldman AG (1986) “Once more on the equilibrium-point hypothesis (λ model) for motor control,” Journal of Motor Behavior 18: 17-54. Feldman AG (1980) Superposition of motor programs – II. Rapid forearm flexion in man. Neuroscience 5:91–95 Feldman AG (1966) “Functional tuning of the nervous system during control of movements of maintenance of a steady posture – II. Controllable parameters of the muscles,” Biophysics 11: 565-578. Fitts PM, Peterson JR (1964) “Information capacity of discrete motor responses,” J Exp Psychol 67: 103-112. Fitts PM (1954) “The information capacity of the human motor system in controlling the amplitude of movement,” J Exp Psychol 47: 381-391. Flanagan JR, Ostry DJ, Feldman AG (1993) “Control of trajectory modifications in target-directed reaching,” Journal of Motor Behavior 25: 140-152. Flanders M, Helms Tillery S, Soechting J (1992) Early stages in a sensorimotor transformation. Behav Brain Sci 15: 309-362 Flash T, Henis E, Inzelberg R, Korczyn AD (1992) “Timing and sequencing of arm trajectories: Normal and abnormal motor behavior,” Human Movement Science 11: 83-100. Flash T, Henis E (1991) “Arm trajectory modifications during reaching towards visual targets,” J Cognitive Neuroscience 3(3): 220-230. Flash T (1990) The organization of human arm trajectory control. In: Winters JM, Woo SL-Y (eds) Multiple muscle systems. Biomechanics and movement organization. Springer, Berlin Heidelberg NewYork, 282–301 Flash, T. (1989) “Generation of reaching movements: Plausibility and implications of the equilibrium trajectory hypothesis,” Brain, Behavior & Evolution, 33(2): 63-8. Flash, T.; Hogan, N. (1985) “The coordination of arm movements: An experimentally confirmed mathematical model,” Journal of Neuroscience, 7: 1688-1703.

153

Flowers K A (1976) “Virtual “open-loop" and "closed-loop" characteristics of voluntary movement in patients with Parkinsonism and intention tremor,” Brain 99: 269-310. Fothergill LC, Griffin MJ (1977) “The subjective magnitude of whole-body vibration,” Ergonomics 20(5): 521-533. Gauthier GM, Pirron JP, Roll JP, Marchetti E, Martin B (1984) “High frequency vestibule-ocular reflex activation through forced head rotation in man,” Aviat Space and Environ Med 55(1): 1-7. Gauthier GM, Roll JP, Martin B, Harlay F (1981) “Effects of whole-body vibration on sensory motor system performance in man,” Aviat Space Environ Med 52(8): 473-479. Gentilucci M, Toni I, Chieffi S, Pavesi G (1994) The role of proprioception in the control of prehension movements: a kinematic study in a peripherally deafferented patient and in normal subjects. Exp Brain Res 99: 483-400 Ghez C, Gordon J, Ghilardi MF (1995) Impairments of reaching movements in patients without proprioception II Effects of visual information on accuracy. J Neurophysiol 73: 361-372 Ghez C (1991) The control of movement. In: Kandel ER, Schwartz JH, Jessell TM (eds) Principles of Neural Science. 3rd Ed. New York, NY: Elsevier Science: 533-547 Ghez C, Gordon J, Ghilardi MF, Christakos CN, Cooper SE (1990) “Roles of proprioceptive input in the programming of arm trajectories,” Cold Spring Harbor Symposium on Quantitative Biology 55: 837-847. Gianaros P, Muth E, Mordkoff J, Levine M, Stern R (2001) A questionnaire for the assessment of the multiple dimensions of motion sickness. Aviation Space and Environmental Medicine 72: 115-119 Gibson JJ (1979) In: The ecological approach to visual perception. Boston: Houghton Mifflin. Goodale M (1988) Modularity in visuomotor control from input to output. In: Robinson T (ed) Behavioral approaches to brain research. New York: Oxford University Press: 41-61 Goodwin AW, McCloskey DI, Matthews PBC (1972) “The contribution of muscle afferents to kinaesthesia shown by vibration induced illusions of movement and by the effects of paralyzing joint afferents,” Brain 95: 705-748. Gordon TJ, Best MC, Dixon PJ (2002) “An automated driver based on convergent vector fields,” Proc Instn Mech Eng 216: 329-347. 154

Gordon J, Ghilardi MF, Ghez C (1995) “Impairments of reaching movements in patients without proprioception. I. Spatial errors,” J Neurophys 73(1): 347-360. Gordon J, Ghilardi MF, Ghez C (1994) “Accuracy of planar reaching movements: Independence of direction and extent variability,” Experimental Brain Research 99: 97–111. Gray, R.; Wilkinson, R.T.; Maslen, K.R.; Rowlands, G.F. (1976) “The effects of 3 hours of vertical vibration at 5 Hz on the performance of some tasks,” Great Britain Ministry of Defence, Royal Aircraft Establishment, Famborough, England/ Medical Research Council, Cambridge, Applied Psychology Research Unit, Report No. RAE-TR-76011. Granata KP, Wilson SE (2001) “Trunk posture and spinal stability,” Clinical Biomech 16(8): 650-659. Gréa H, Desmurget M, Prablanc C (2000) “Postural invariance in threedimensional reaching and grasping movements,” Experimental Brain Research 134: 155-162. Gribble P, Mullin L, Cothros N, Mattar A (2003) Role of cocontraction in arm movement accuracy. J Neurophysiol 89: 2396-2405. Griffin MJ (2001) “The validation of biodynamic models,” Clinical Biomechanics 16(Supplement No.1): S81-S92. Griffin MJ (1990) Handbook of human vibration. San Diego: Academic Press. Griffin MJ (1986) “Evaluation of vibration with respect to human response,” SAE International Congress and Exposition, Warrendale, PA, 11-34. Griffin MJ, Whitman EM (1978) “Individual variability and its effect on subjective and biodynamic response to whole-body vibration,” Journal of Sound and Vibration 58: 239-250. Hagena FW, Wirth CJ, Piehler J, Plitz W, Hofmann GO, Zwingers T (1985) “Invivo experiments on the response of the human spine to sinusoidal Gz-vibration,” AGARD Conference Proceedings 378(16): 1-12. Harris, C.S.; Shoenberger, R.W. (1966) “Effects of frequency of vibration on human performance,” Journal of Engineering Psychology, 5(1): 1-15. Heath (2005) Role of limb and target vision in the online control of memoryguided reaches. Mot Cont 9: 281-311 Hicks S (1960) The effects of eight hours confinement in mobile armored personnel carriers on selective combat relevant skills: Study II. (Rep No TM 1760) Aberdeen Proving Ground, MD: Human Engineering Laboratory

155

Hinz B, Seidel H, Bräuer D, Menzel G, Blüthner R, Erdmann U (1988) “Bidimensional accelerations of lumbar vertebrae and estimation of internal spinal load during sinusoidal vertical whole-body vibration: a pilot study,” Clinical Biomechanics 3: 241-248. Hogan N (1984) “Adaptive control of mechanical impedance by coactivation of antagonist muscles,” IEEE Transactions on Automatic Control 29: 681-690. Howarth CI, Beggs WDA, Bowden JM (1971) “The relationship between speed and accuracy of movement aimed at a target,” Acta Psychologica 35: 207-218. Hsiang SM, Ayoub MM (1994) “Development of methodology in biomechanical simulation of manual lifting,” Int J Indust Ergon 19: 59–74. Ihaka R., Gentleman R. (1996) "R: A language for data analysis and graphics," J. of Computational and Graphical Statistics 5: 299-314. International Organization for Standardization (1997) ISO 2631-1 (E). Mechanical vibration and shock: Evaluation of human exposure to whole-body vibration. Part 1: General requirements. Jagacinski RJ, Flach JM (2003) In: Control Theory for Humans. Lawrence Erlbaum Associates. Jagacinski, R. J. (1989) “Target acquisition: Performance measures, process models, and design implications,” Applications on human performance models to system design, 135-149. Jagacinski RJ, Monk DL (1985) “Fitts' law in two dimensions with hand and head movements,” J Motor Behavior 17: 77-95. Jagacinski RJ, Repperger DW, Moran MS, Ward SL Glass B (1980) “Fitts' law and the microstructure of rapid discrete movements,” Journal of Experimental Psychology: Human Perception and Performance 6(2): 309-320. Jalzem PF, Gledhill RB (1993) “Predicting height from arm measurements,” J Pediatr Orthop 13(6): 761-765. Jeannerod M (1991) A neurophysiological model for the directional coding of reaching movements. In: Paillard J (ed) Brain and space. Oxford: Oxford University Press, 49-69 Jeannerod M (1988) In: The neural and behavioural organization of goal directed movements. Clarendon Press, Oxford. Jex HR, Magdaleno RE (1978) “Progress in measuring and modeling the effects of low frequency vibration on performance,” In: von Gierke HE (ed) AGARD

156

Conference Proceedings CP-253: Models and analogues for the evaluation of human biodynamic response, performance and protection. Paris, France. Johanning E (1991) “Survey results of back disorders and health problems in subway train operators exposed to whole-body vibration,” Scandinavian Journal of Work Environment and Health 17: 414-419. Junker, A.; Levison, W. (1980) “Some empirical techniques for human operator performance measurement,” Proceedings of the International Conference of the Cybernetics Society, 101-105. Kagerer FA, Contreras-Vidal JL, Stelmach GE (1997) “Adaptation to gradual as compared with sudden visuo-motor distortions,” Exp Brain Res 115: 557-561. Karapetsas A, Vlachos F (1997) ex and handedness in development of visuomotor skills. Perceptual and Motor Skills 85: 131-140 Kauranen K, Vanharanta H (1996) Influences of aging, gender, and handedness on motor performance of upper and lower extremities. Perceptual and Motor Skills 82 515-525 Keele S, Posner M (1968) “Processing of visual feedback in rapid movements,” J Exp Psych 77: 155-158. Kelso JAS (1982) “Concepts and issues in human motor behavior: coming to grips with the jargon,” In: Kelso JAS (ed) Human motor behavior: an introduction. Hillsdale, NJ: Erlbaum. Kerr BA, Langolf GD (1977) “Speed of aiming movements,” Quarterly Journal of Experimental Psychology, 29: 475-481. Kim KH, Martin BJ (2005) “Postural goal of unconstrained head movement strategies,” 35th Annual Meeting of the Society for Neuroscience, Washington, DC. Kitazaki S, Griffin MJ (1998) “Resonance behaviour of the seated human body and effects of posture,” Journal of Biomechanics 31: 143–149. Kleinman D, Baron S, Levison W (1971) “Control theoretic approach to mannedvehicle systems analysis,” IEEE Trans Autom Control, AC-16 (6): 824-832. Lacquaniti F, Terzuolo C, Viviani P (1983) “The law relating the kinematic and figural aspects of drawing movements,” Acta Psychol 54: 115-130 Langolf, G. D.; Chaffin, D. B.; Foulke, J. A. (1976) “An investigation of Fitts' law using a wide range of movement amplitudes,” Journal of Motor Behavior, 8: 113128.

157

Lashley KS (1951) The problem of serial order in behavior. In: Jeffress LA (ed) Cerebral mechanisms in behavior: The Hixon symposium. New York: Wiley. Lewis CH, Griffin MJ (1979) “Mechanisms of the effects of vibration frequency, level and duration on continuous manual control performance,” Ergonomics 22(7): 855-889. Lewis CH, Griffin MJ (1978a) “Predicting the Effects of Dual-Frequency Vertical Vibration on Continuous Manual Control Performance,” Ergonomics 21(8): 637650. Lewis CH, Griffin MJ (1978b) “A review of the effects of vibration on visual acuity and continuous manual control. Part II: Continuous manual control,” Journal of Sound and Vibration 56(3): 415-457. Lewis CH, Griffin MJ (1976) “The effects of vibration on manual control performance,” Ergonomics 19: 203-216. Lewis J (1962) “A partial review of the literature on physiological disorders resulting from the operations of motor vehicles,” (Rep No TM 17-62) Aberdeen Proving Ground, MD: Human Engineering Laboratory. Li G, Baker SP, Grabowski JG, Rebok GW (2001) “Factors associated with pilot error in aviation crashes,” Aviation, Space and Environmental Medicine 72: 5258. Liang C-Y, Magdaleno R, Lee D, Klyde DH, Allen RW, Rider K, Overmeyer K (2005) “A biodynamic model for the assessment of human operator performance under vibration environment,” Technical Paper 2005-01-2742. SAE International, Warrendale, PA. Lins, W.; Dugoff, H. (1972) “Motion simulation and its application to ride dynamics research,” SAE Paper 720003. Llyod DCEF, Troup JDG (1983) “Recurrent back pain and its prediction,” Journal of Occupational Medicine 33: 66-74. Luce R (1986) Response time: their role in inferring elementary mental organization. Oxford: Oxford University Press MacAdam CC (1980) “An optimal preview control for linear systems,” J Dynamic Systems, Measurement, and Control, September, 188-193. MacKenzie IS (1995) Movement time prediction in human-computer interfaces. In: Baecker RM, Buxton WAS, Grudin J, Greenberg S (eds) Readings in humancomputer interaction. 2nd ed: 483-493.

158

MacKenzie IS (1992) “Fitts' law as a research and design tool in humancomputer interaction,” Human-Computer Interaction 7: 91-139. MacKenzie IS, Buxton W (1992) “Extending Fitts’ law to two-dimensional tasks,” In: Proceedings of the CHI conference on human factors in computing systems. ACM press: 219-226. MacKenzie CI, Mateniuk RG, Dugas C, Liske D, Eickmeier B (1987) “Threedimensional movement trajectories in Fitts' task: Implications for control,” Quarterly Journal of Experimental Psychology 39A: 629-647. Mansfield NJ, Holmlund P, Lundstrom R, Lenzuni P, Nataleteti P (in press, 2006) “Effect of vibration magnitude, vibration spectrum and muscle tension on apparent mass and cross axis transfer functions during whole-body vibration exposure,” J Biomechanics. Mansfield NJ (2005) “Impedance methods (apparent mass, driving point mechanical impedance and absorbed power) for assessment of the biomechanical response of the seated person to whole-body vibration,” Industrial Health 43: 378-389. Mansfield N (2005) Human Response to Vibration. CRC Press, Boca Raton, Florida. Mansfield, N.J.; Griffin, M.J. (2000) “Non-linearities in apparent mass and transmissibility during exposure to whole-body vertical vibration,” Journal of Biomechanics, 33: 933-941. Mars F, Archambault PS, Feldman AG (2003) “Vestibular contribution to combines arm and trunk motion,” Exp Brain Res 150: 515-519. Marteniuk RG, MacKenzie CI, Jeannerod M, Athens S, Dugas C (1987) “Constraints on human arm movement trajectories,” Canadian Journal of Psychology 41(3): 365-378. Martin BJ, Park H (1997) Analysis of the tonic vibration reflex: influence of vibration variables on motor unit synchronization and fatigue. Eur J Appl Physiol 75: 504-511 Martin BJ, Saltuman J, Elders G (1997) “Effects of vibration frequency and duration on eye-hand coordination in pointing tasks,” Proceedings of the 19th Annual International Conference of the IEEE 6(6): 2793-2798. Martin BJ, Roll JP, DiRenzo N (1991) “The Interaction of Hand Vibration with Oculo-manual Coordination in Pursuit Tracking,” Aviat Space and Environ Med 62(2): 145-152.

159

Martin BJ, Roll JP, Gauthier GM (1984) “Spinal reflex alterations as a function of intensity and frequency of vibration applied to the feet of seated subjects,” Aviation, Space and Environmental Medicine 55: 8-12. Matsumoto Y, Griffin MJ (2001) “Modelling the dynamic mechanisms associated with the principal resonance of the seated human body,” Clinical Biomechanics 16(Supplement No 1): S31-S44. Matsumoto, Y.; Griffin, M.J. (2000) “Comparison of biodynamic responses in standing and seated human bodies,” Journal of Sound and Vibration, 238(4): 691-704. Matsumoto, Y.; Griffin, M.J. (1998) “Dynamic response of the standing human body exposed to vertical vibration: Influence of posture and vibration magnitude,” Journal of Sound and Vibration, 212: 85-107. Matthews PBC (1972) In: Mammalian muscle receptors and their central actions. London: Arnold. McDowell K, Jeka J, Schöner G, Hatfield B (2002) Behavioral and electrocortical evidence of an interaction between probability and task metrics in movement preparation. Exp Brain Res 144: 303-313 McLeod RW, Griffin MJ (1989) “A Review of the Effects of Translational WholeBody Vibration on Continuous Manual Control Performance,” Journal of Sound and Vibration 133(1): 55-115. McLeod RW, Griffin MJ (1988) “Performance of a complex manual control task during exposure to vertical whole-body vibration between 0.5 and 5.0 Hz,” Ergonomics 31(8): 1193-1203. McLeod P, Poulton C, Du Ross H, Lewis W (1980) The influence of ship motion on manual control skills. Ergonomics 23: 623-634 McRuer D (1980) “Human dynamics in man-machine systems,” Automatica 16: 237-253. McRuer DT, Graham D, Krendel ES, Reisener Jr W (1965) “Human pilot dynamics in compensatory systems: Theory, models, and experiments with controlled element and forcing function variations,” (AFFRDL-TR- 65-15). WrightPatterson AFB, OH: Air Force Flight Dynamics Laboratory. McRuer DT, Krendel ES (1959 “The human operator as a servo system element,” J Franklin Institute 267: 381-403. Messenger AJ, Griffin MJ (1989) “Effects of Anthropometric and Postural Variables on the Transmission of Whole-Body Vertical Vibration from Seat-To-

160

Head,” Southampton University, Institute of Sound and Vibration Research, Technical Report No. 172. Messier J, Adamovich S, Berkinblit M, Tunik E, Poizner H (2003) Influence of movement speed on accuracy and coordination of reaching movements to memorized targets in three-dimensional space in a deafferented subject. Exp Brain Res 150: 399-416 Messier J, Kalaska J (1999) “Comparison of variability of initial kinematics and endpoints of reaching movements,” Experimental Brain Research 125: 139-152. Messier J, Kalaska J (1997) “Differential effect of task conditions on errors of direction and extent of reaching movements,” Experimental Brain Research 115: 469-478. Meyer D, Smith J, Kornblum S, Abrams R, Wright C (1990) “Speed-accuracy tradeoffs in aimed movements: Toward a theory of rapid voluntary action,” In: Jeannerod M (ed), Attention and performance XIII. Hillsdale, NJ: Erlbaum, 173226. Meyer D, Abrams R, Kornblum S, Wright C, Smith J (1988) “Optimality in human motor performance: Ideal control of rapid aimed movements,” Psychological Review 95(3): 340-370. Meyer D, Smith J, Wright C (1982) “Models for the speed and accuracy of aimed limb movements,” Psychological Review 89: 449-482. Milner T (2002) Adaptation to destabilizing dynamics by means of muscle cocontraction. Exp Brain Res 143: 406-416 Milner TE, Cloutier C (1998) “Damping of the wrist joint during voluntary movement,” Experimental Brain Research 122: 309-317. Milner TE, Cloutier C (1993) “Compensation for mechanically unstable loading in voluntary wrist movement,” Experimental Brain Research 94: 522-532. Miwa T (1968) “Evaluation methods for vibration effect, Part 4. Measurements of vibration greatness for whole body and hand in vertical and horizontal vibrations,” Industrial Health 6: 1-10. Morasso P (1983) “Three-dimensional arm trajectories,” Biological Cybernetics 48: 187-194. Morasso P, Mussa Ivaldi FA (1982) “Trajectory formation and handwriting: A computational model,” Biological Cybernetics 45: 131-142. Morasso P (1981) “Spatial control of arm movements,” Experimental Brain Research 42: 223-227.

161

Mori H, Hoshino J (2002) “Independent component analysis and synthesis of human motion,” Proceedings of the 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing 4: 3564-3567. Moseley M, Griffin M (1986) “Effects of display vibration and whole-body vibration on visual performance,” Ergonomics 29: 977-983. Murata A, Iwase H (2001) “Extending Fitts’ law to a three-dimensional task,” Human Movement Science 20: 791-805. Newell KM, Carlton LG, Hancock PA (1984) Kinetic analysis of response variability. Psychol Bull 96: 133–151 Newell, K.; Carlton, L.; Carlton, M. (1980) “Velocity as a factor in movement timing accuracy,” Journal of Motor Behavior, 12: 47-56. Nichols TR, Houk JC (1976) “Improvement in linearity and regulation of stiffness that results from actions of stretch reflex,” Journal of Neurophysiology 39: 119-142. Nicol, J.; Morrison, J.; Roddan, G.; Rawicz, A. (1997) “Modelling the dynamic response of the human spine to shock and vibration using a recurrent neural network,” Heavy Vehicle Systems, 4(2-4): 145-165. Oborne, D.; Boarer, P. (1982), “Subjective response to whole-body vibration – the effects of posture,” Ergonomics, 25(7): 673-681. Okadome, T.; Honda, M. (1999) “Kinematic construction of the trajectory of sequential arm movements,” Biological Cybernetics, 80: 157-169. Paddan GS, Griffin MJ (1993) “Transmission of vibration through the human body to the head: a summary of experimental data,” Institute of Sound and Vibration Research, Technical Report 218. Paillard J (1996) “Fast and slow feedback loops for the visual correction of spatial errors in a pointing task: a reappraisal,” Canadian Journal of Physiol Pharmacol 74: 401-417. Parks, D.L. (1961) “A comparison of sinusoidal and random vibration effects on human performance,” Boeing Company, Human Factors Unit, Wichita, Kan., Technical Report No. 2. Peterson JR, Fitts PM (1964) “Information capacity of discrete motor responses,” Journal of Experimental Psychology 67: 103-112. Pipenbrinck N (2006) (cubic.org/docs/hermite.htm).

“Hermite

162

Curve

Interpolation,”

Plamondon R (1998) “A kinematic theory of rapid human movements: Part III. Kinetic outcomes,” Biological Cybernetics, 78: 133-145. Plamondon R (1997) “Speed/accuracy trade-offs in target-directed movements,” Behavioral and Brain Sciences 20: 279-303. Plamondon R (1995a) “A kinematic theory of rapid human movements: Part I. Movement representation and generation,” Biological Cybernetics 72: 295-307. Plamondon R (1995b) “A kinematic theory of rapid human movements: Part II. Movement time and control,” Biological Cybernetics 72: 309-320. Plamondon R, Alimi AM (1997) “Speed/accuracy trade-offs in target-directed movements,” Behavioral and Brain Science 20(2): 279-303. Plamondon R (1991) “On the origin of asymmetric bell-shaped velocity profiles,” Tutorials in Motor Neuroscience, 283-295. Plamondon R (1990) “A unified approach to the study of target directed movements,” NATO Advanced Study Institute, Tutorials in Motor Neuroscience, 15-24. Polit A, Bizzi E (1979) “Characteristics of motor programs underlying arm movements in monkeys,” Journal of Neurophysiology 42: 183-194. Polit A, Bizzi E (1978) “Processes controlling arm movements in monkeys,” Science 201: 1235-1237. Pope MH, Kaigle AM, Magnusson M, Broman H, Hansson T (1991) “Intervertebral motion during vibration,” Journal of Engineering in Medicine 205: 39-44. Pope M, Wilder D, Jorneus L, Broman H, Svensson M, Andersson G (1987) “Response of the seated human to sinusoidal vibration and impact,” Journal of Biomechanical Engineering 109(4): 279-284. Pope M, Wilder D, Jorneus L, Broman H, Svensson M, Andersson G (1986) “Response of the seated human to vibration and impact,” Proceedings of the 12th Annual Northeast Bioengineering Conference, IEEE: 31-34. Prablanc C, Echallier J, Jeannerod M, Komilis E (1979) Optimal response of eye and hand motor systems in pointing at a visual target I: Spatio-temporal characteristics of eye and hand movements and their relationships when varying the amount of visual information. Biol Cybern 35: 183-187. Prokop G (2001) “Modeling human vehicle driving by model predictive online optimization,” Vehicle System Dynamics 35: 19-53.

163

Psotka J, Lewis SA, King D (1998) “Effects of field of view on judgments of selflocation: Distortions in distance estimations even when the image geometry exactly fits the field of view,” Presence: Teleoperators and Virtual Environments 7(4): 352-369. Ranganathan, R.; Mohan, M. (1997) “Review of the general effects of wholebody vibration,” Heavy Vehicle Systems, 4(2-4): 353-372. Rao, B.; Jones, B. (1978) “Equal sensation study of seated subjects in three translational modes,” Ergonomics, 21(2): 123-134. Reason J, Brand J (1975) Motion sickness. London: Academic Press Redfern MS, Muller MLTM, Jennings JR, Furman JM (2002) “Attentional dynamics in postural control during perturbations in young and older adults,” J Gerontology Series A: Biological Sciences and Medical Sciences 57: B298-B303. Reed MP, Parkinson MB, Chaffin DB (2003) “A new approach to modeling driver reach,” Technical Paper 2003-01-0587. SAE International, Warrendale, PA. Richardson, M.; Flash, T. (2002) “Comparing smooth arm movements with the two-thirds power law and the related segmented-control hypothesis,” Journal of Neuroscience, 22(18): 8201-8211. Rider K, Martin B (2005) “Superposition of optimal submovements in feedbackcontrolled reaching,” XXth Congress of the International Society of Biomechanics, Cleveland, Ohio. Rider K, Martin B (2004) Degradation of exocentric reference in visually-occluded three-dimensional reaching tasks? 34th Annual Meeting of the Society for Neuroscience, San Diego, CA Rider K, Chaffin D, Nebel K, Mikol K (2004) Modeling in-vehicle reaches perturbed by ride motion. SAE Transactions: J Aerospace 113(1): 193-198 Rider K, Chaffin D, Nebel K, Mikol K, Reed M (2003) A pilot study of the effects of vertical ride motion on reach kinematics. SAE Transactions: J Passenger Cars 112(6): 719-725 Riemann B, Lephart S (2002) “The sensorimotor system. Part II: The role of proprioception in motor control and functional joint stability,” J Athletic Training 37(1): 80-84. Roll JP, Vedel JP (1982) “Kinaesthetic role of muscle afferents in man, studied by tendon vibration and microneurography,” Experimental Brain Research 47(2): 177-190.

164

Rolland-Cachera MF, Cole TJ, Sempé M, Tichet J, Rossignol C, Charraud A (1991) “Body mass index variations: centiles from birth to 87 years,” European Journal of Clinical Nutrition 45: 13-21. Rosenbaum D.A. (1980) “Human movement initiation: specification of arm, direction, and extent,” Journal of Experimental Psychology 109(4): 444–474. Rosetti Y, Desmurget M, Prablanc C (1995) Vectorial Coding of Movement: Vision, Proprioception, or Both? J Neurosci 74(1): 457-463 Rowlands, G.F. (1977) “The transmission of vertical vibration of the heads and shoulders of seated men,” Great Britain Ministry of Defence, Royal Aircraft Establishment, Famborough, England. Published by HMSO, London, England, Report No. TR 77068. Rudisill M, Toole T (1993) Gender differences in motor performance of 50- to 79year-old adults. Perceptual and Motor Skills 77(3, pt 1): 939-947 Sabes P (2000) “The planning and control of reaching movements,” Current Opinion in Neurobiology 10: 740-746. SAFEWORK (2006) “The SAFEWORK advantage,” SAFEWORK Inc. Human Modeling Technology. http://www.safework.com. Sainburg R (2002) Evidence for a dynamic-dominance handedness. Exp Brain Res 142: 241-258

hypothesis of

Sandover J (1983) “Dynamic loading as a possible source of low-back disorders,” Spine 8: 652-658. Sanger T (2000) Human arm movements described by a low-dimensional superposition of principal components. J Neurosci 20(3): 1066-1072 Sarlegna F, Blouin J, Vercher J-L, Bresciani J-P, Bourdin C, Gauthier G (2004) Online control of the direction of rapid reaching movements. Experimental Brain Research 157: 468-471 Saunders J, Knill D (2004) Visual feedback control of hand movements. J Neurosci 24(13): 3223-3234 Schaal S (2002) Arm and hand movement control In: Arbib MA (ed) The handbook of brain theory and neural networks. 2nd ed. Cambridge, MA: MIT Press, 110-113 Schaal, S.; Sternad, D. (2001) “Origins and violations of the 2/3 power law in rhythmic three-dimensional arm movements,” Experimental Brain Research, 136: 60-72.

165

Schipani S, Bruno R, Lattin M, King B, Patton D (1998) “Quantification of cognitive process degradation while mobile, attributable to the environmental stressors endurance, vibration, and noise,” (Rep No ARL-TR-1603) Aberdeen Proving Grounds, MD: Army Research Laboratory. Schmidt R, Lee T (2005) In: Motor control and learning: a behavioral emphasis. 4th ed. Champaign: Human Kinetics. Schmidt R.A.; Sherwood D.; Zelaznik H.N.; Leikind B. (1985) “Speed-accuracy tradeoffs in motor behavior: Theories of impulse variability,” Motor Behavior: Programming, Control, and Acquisition, 79-123. Schmidt RA (1980) “Past and future issues in motor programming,” Research Quarterly for Exercise and Sport 51: 122-140. Schmidt RA, Zelaznik HN, Hawkins B, Franck JS, Quinn JT (1979) “Motor output variability: A theory for accuracy of rapid motor acts,” Psychological Review 86: 415-451. Schmidt RA, Zelaznik HN, Frank JS (1978) Sources of inaccuracy in rapid movement, In: Information processing in motor control and learning, 183-203. Schneider GE (1969) “Two visual systems,” Science 163: 895-902. Scholz JP, Schöner G (1999) “The uncontrolled manifold concept: identifying control variables for a functional task,” Exp Brain Res 126: 289-306. Sergio L, Scott S (1998) Hand and joint paths during reaching movements with and without vision. Exp Brain Res 122: 157-164 Servos P (2000) Distance estimation in the visual and visuomotor systems. Exp Brain Res 130: 35-47 Shneiderman B (1992) In: Designing the user interface: strategies for effective human-computer interaction. Boston, MA: Addison-Wesley Longman Publishing. Shöenberger (1974) “Mechanisms of vibration effects on aircrew performance,” AGARD Conference Proceedings on Vibration and Combined Stresses in Advanced Systems, Aerospace Medical Panel Specialists’ Meeting, Oslo, Norway, Paper B-17, Advisory Group for Aerospace Research and Development. Soechting J, Flanders M (1992) Moving in three-dimensional space: frames of reference, vectors, and coordinate systems. Annual Review in Neuroscience 15: 167-191 Soechting JF, Flanders M (1991) “Deducing central algorithms of arm movement control from kinematics,” In: Motor Control: Concepts and Issues, eds., D.R. Humphrey and H.-J. Freund. John Wiley & Sons Ltd., 293-306.

166

Soechting JF, Flanders M (1989a) “Sensorimotor representations for pointing to targets in three-dimensional space,” Journal of Neurophysiology 62(2): 582-594. Soechting JF, Flanders M (1989b) “Errors in pointing are due to approximations in sensorimotor transformations,” Journal of Neurophysiology 62(2): 595-604. Soechting JF, Lacquaniti F (1983) “Modification of trajectory of a pointing movement in response to a change in target location,” Journal of Neurophysiology 49: 548-564. Soechting J, Lacquaniti F (1981) “Invariant characteristics of a pointing movement in man,” Journal of Neuroscience 1(7): 710-720. Spijkers, W. A. C.; Sanders, A. F. (1984) “Spatial accuracy and programming of movement velocity,” Bulletin of the Psychonomic Society, 22(6): 531-534. Staines WR, McIlroy W, Brooke J (2001) Cortical representation of whole-body movement is modulated by proprioceptive discharge in humans. Exp Brain Res 138: 235-242 Sternad D, Schaal S (1999) “Segmentation of endpoint trajectories does not imply segmented control,” Experimental Brain Research 124: 118-136. Sutjiatmo, B. (1991) “Human perception response of vehicle excited by random road roughness,” Proceedings of the 6th International Pacific Conference on Automotive Engineering, 1165-1172. Suzuki, M.; Yamazaki, Y.; Mizuno, N.; Matsunami K. (1997) “Trajectory formation of the center-of-mass of the arm during reaching movements,” Neuroscience, 76(2): 597-610. Todorov E (2004) Optimality principles in sensorimotor control. Nat Neurosci 7:907-915 Unigraphics Solutions (2006) “Technomatix Jack: digital human model”, UGS PLM-Solutions/Technomatix. http://www.ugs.com. Uno Y, Kawato M, Suzuki R (1989) Formation and control of optimal trajectory in human multijoint arm movement: Minimum torque-change model. Biol. Cybern. 61: 89-101 Van Beers R, Sittig A, van der Goin J (1996) Localization of a seen finger is based exclusively on proprioception and visual position information. Exp Brain Res 111: 253-261 Viviani P, Flash T (1995) “Minimum-jerk, two-thirds power law and isochrony: Converging approaches to the study of movement planning,” Journal of Experimental Psychology: Perception and Performance 21: 32-53.

167

Viviani P, Terzuolo C (1982) “Trajectory determines movement dynamics,” Neuroscience 7: 431-437. Vogt LE, Coermann RR, Fust HD (1968) “Mechanical impedance of the sitting human under sustained acceleration,” Aerospace Medicine 39: 675-679. von Gierke H, Brammer A (1996) Effects of shock and vibration on humans. In: Harris, CM (ed) Shock and Vibration Handbook. 4th Ed. McGraw-Hill Wade MG, Newell KM, Wallace SA (1978) “Decision time and movement time as a function of response complexity in retarded persons,” American Journal of Mental Deficiency 83: 135-144. Wakeling JM, Nigg BM, Rozitis AI (2002) “Muscle activity damps the soft tissue resonance that occurs in response to pulsed and continuous vibrations,” J Appl Physiol 93: 1093-1103. Wasserman, D. (1991) “Human Vibration Standards,” Journal of Sound and Vibration, July 1991: 30-32. Welford AT (1968) In: Fundamentals of skill. London: Methuen. Werner EB (1991) In: Manual of visual fields. New York: Churchill Livingstone. Westwood D, Heath M, Roy E (2003) No evidence for accurate visuomotor memory: Systematic and variable error in memory-guided reaching. Journal of Motor Behavior 35(2): 127-133 Whitman EM, Griffin MJ (1978) “The effects of vibration frequency and direction on the locations of areas of discomfort caused by whole-body vibration,” Applied Ergonomics 9: 231-239. Wierwille WW, Tijerina L (1996) “An analysis of driving accident narratives as a means of determining problems caused by in-vehicle visual allocation and visual workload,” In: Vision in Vehicles (Gale AG et al., eds). Amsterdam: NorthHolland. Wiker SF, Langolf GD, Chaffin DB (1989) “Arm posture and human movement capability,” Human Factors 31: 421-441. Wikipedia (2006) “Cubic Hermite spline,” In: Wikipedia, The Free Encyclopedia. http://en.wikipedia.org/wiki/Cubic_Hermite_spline. Wilder DG, Woodworth BB, Frymoyer JW, Pope MH (1982) “Vibration and the human spine,” Spine 7: 243-254. Woldstad J, Ayoub M (1998) “A model to predict human motion during lifting,” SAE Digital Human Modeling Conference, Dayton, OH. Tech Paper #981303.

168

Wolpert D, Ghahramani Z (2000) “Computational principles of movement neuroscience,” Nat Neurosci 3(Suppl): 1212-1217. Wolpert D, Ghahramani Z, Jordan M (1995) “An internal model for sensorimotor integration,” Science 269: 1880-1882. Woodson WE, Tillman B, Tillman P (1992) In: Human Factors Design Handbook (2nd Ed). McGraw-Hill, Inc. New York. Woodworth RS (1899) “The accuracy of voluntary movement,” Psychological Review, 3: 1-114. Zelaznik H, Mone S, McCabe G, Thaman C (1988) “Role of temporal and spatial precision in determining the nature of the speed-accuracy trade-off in aimedhand movements,” Journal of Experimental Psychology 14(2): 221-230. Zelaznik HN, Schmidt RA, Gielen S (1986a) “Kinematic properties of aimed hand movements,” Journal of Motor Behavior, 18: 353-372. Zelaznik HN, Schmidt RA, Gielen S (1986b) “Kinematic properties of rapid aimed hand movements,” Journal of Motor Behavior, 18: 353-373. Zelaznik H, Hawkins B, Kisselburgh L (1983) “Rapid visual feedback processing in single-aiming movements,” Journal of Motor Behavior 15: 217-236. Zelaznik HN, Shapiro DC, McColsky D (1981) “Effects of a secondary task on the accuracy of single aiming movements,” Journal of Experimental Psychology: Human Perception and Performance, 7: 1007-1018. Zhang X (2001) “Biomechanical realism versus algorithmic efficiency: a trade-off in human motion simulation modeling,” Technical Paper 2001-01-2090. SAE International, Warrendale, PA. Zhang X, Chaffin D (2000) “A three-dimensional dynamic posture prediction model for in-vehicle seated reaching movements: development and validation,” Ergonomics 43(9): 1314-1330. Zwahlen HT (1993) “Evaluation of pushbutton arrangements in automobiles,” Proceedings of the 37th Annual Meeting of the Human Factors and Ergonomics Society, 2: 969-973.

169

EFFECTS OF RIDE MOTION PERTURBATION ON THE ...

Automotive Research Center (ARC), a research partnership between the. University of ... me during these years, but I plan to spend the rest of my life trying. iv ... Chapter 4 - Study 2: Analysis of the effects of vibration frequency, magnitude, ..... degree of freedom (6DOF) ride motion on the speed and accuracy of in-vehicle.

4MB Sizes 2 Downloads 268 Views

Recommend Documents

Effects of Vertical Land-Vehicle Ride Motion on Human ...
Effects of Vertical Land-Vehicle Ride Motion on Human Reach Performance and. Perceived Task Difficulty. Kevin A. Rider, Woojin Park, and Michael P. Prohaska. Proceedings of the 47th Annual Meeting of the Human Factors and Ergonomics Society,. Denver,

ON A PERTURBATION DETERMINANT FOR ...
OPERATORS. KONSTANTIN A. MAKAROV, ANNA SKRIPKA∗, AND MAXIM ZINCHENKO†. Abstract. For a purely imaginary sign-definite perturbation of a self-adjoint operator, we obtain exponential representations for the perturbation determinant in both upper an

Re-ranking Search Results based on Perturbation of ...
Apr 29, 2006 - search results. ▫. Perturb the graph and see if the resulting graph is still relevant to the query. ▫. Re-rank the search results based on the amount of perturbation .... Microsoft monopoly. Globalization and Democracy. Recent Eart

The Effects of The Inflation Targeting on the Current Account
how the current account behaves after a country adopts inflation targeting. Moreover, I account for global shocks such as US growth rate, global real interest rate ...

Effects of sample size on the performance of ... -
area under the receiver operating characteristic curve (AUC). With decreasing ..... balances errors of commission (Anderson et al., 2002); (11) LIVES: based on ...

Effects of Bending Excitation on the Reaction of ...
Mar 14, 2005 - on abstraction reactions because energy is placed directly into .... absorption spectrum at 300 K from the HITRAN database[21] convo-.

Effects of chemical synapses on the enhancement of ...
where b=0.45, B1 =0.05; CC, gsyn=0.15; EC, gsyn=0.1. EFFECTS OF CHEMICAL SYNAPSES ON THE… PHYSICAL REVIEW E 76, 041902 (2007). 041902-3 ...

the effects of turbidity on perception of risk
Aug 17, 2011 - http://rsbl.royalsocietypublishing.org/content/early/2011/08/03/rsbl.2011.0645.full.html. This article cites 16 articles, 2 of which can be accessed ...

Report of Current Research on the Effects of Second Language ...
example in the United States, and in immersion programmes, as in Canada. We will concentrate here on the Canadian .... United States and found that students who were taught in their first language while receiving intensive instruction in English ....

Effects of chemical synapses on the enhancement of ...
School of Electrical Engineering and Automation, Tianjin University, Tianjin 300072, People's Republic of ..... These lead to the decrease of SR in both of the two.

Effects of disturbance on the reproductive potential of ...
took place in December 1993 and data in the present study pertains to ... using a line-intercept method. ..... ous effect for seedlings being often light interception.

Product Market Evidence on the Employment Effects of the Minimum ...
Apr 4, 2006 - factors, the elasticity of labor supply, and the elasticity of product demand. ... workers at or near the minimum, accounting for roughly a fifth of ...

The effects of sharing attentional resources on the ...
marking in second language acquisition. In T. Huebner & C.A. Ferguson (Eds.), .... Dvorak, Trisha; & Lee, James, eds. Foreign Language Learning: A. Research ...

The effects of increasing memory load on the ... - Springer Link
Apr 27, 2004 - Abstract The directional accuracy of pointing arm movements to remembered targets in conditions of increasing memory load was investigated using a modified version of the Sternberg's context-recall memory-scanning task. Series of 2, 3

Modeling the Effects of Dopamine on the Antisaccade ... - Springer Link
excitation and remote inhibition. A saccade was initiated when ..... Conference of Hellenic Society for Neuroscience, Patra, Greece (2005). [7] Kahramanoglou, I.

8-the effects of the tourist's expenditure on malaysia economy.pdf ...
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. 8-the effects of ...

B003 The Effects Of The Airline Deregulation On Shareholders ...
B003 The Effects Of The Airline Deregulation On Shareholders' Wealth.pdf. B003 The Effects Of The Airline Deregulation On Shareholders' Wealth.pdf. Open.

Effects of the Edm Combined Ultrasonic Vibration on the Machining ...
A cylindrical copper tungsten bar was used as the electrode material for .... of the Edm Combined Ultrasonic Vibration on the Machining Properties of Si3N4.pdf.

Global-motion aftereffect does not depend on awareness of the ...
Jan 16, 2014 - ly spaced over the 360º range. Observers were .... integer division by 360. ..... awareness while allowing observers to view the same motion.

Global-motion aftereffect does not depend on awareness of the ...
Jan 16, 2014 - Abstract It has been shown that humans cannot perceive more than three directions from a multidirectional motion stimulus. However ...

anthropogenic effects on population genetics of ... - BioOne
6E-mail: [email protected] ... domesticated status of the host plant on genetic differentiation in the bean beetle Acanthoscelides obvelatus.

EFFECTS OF SURFACE CATALYTICITY ON ...
The risk involved, due to an inadequate knowledge of real gas effects, ... the heat shield surface, increase the overall heat flux up to about two times, or more, higher than ..... using data from wind tunnel and free flight experimental analyses.

review of the literature on the biological effects of wireless radiaton on ...
Nov 3, 2014 - E. Mammals: contact M. Friesen (email listed above) for the 1,000+ reference list. ... domesticus): A Possible Link with Electromagnetic Radiation. Electromagnetic. Biology and ... occurred. After a recovery period, the ants were able t