©2007 Emerald Group. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the Emerald Group.

Research article

Path-planning techniques for the simulation of disassembly tasks Iker Aguinaga CEIT, Donostia - San Sebastia´n, Gipuzkoa, Spain, and

Diego Borro and Luis Matey CEIT, Donostia - San Sebastia´n, Gipuzkoa, Spain and TECNUN, Donostia-San Sebastia´n, Gipuzkoa, Spain Abstract Purpose – This paper aims to develop path-planning techniques that support a general selective disassembly planner in a virtual reality environment. Design/methodology/approach – The paper presents an automatic selective disassembly planning and two path-planning techniques that support it. The first one is based on single translations, while the second is based on the generation of a random search tree. The methods used have been adapted and modified from available robotic path-planning methods for their use in disassembly path planning. Findings – The paper finds that the proposed techniques are applicable to the automatic generation of disassembly sequences. Research limitations/implications – The paper provides an automatic tool that can be integrated in simulation software for the analysis and validation of disassembly operation. Practical implications – Maintenance operations have a great impact in the security and life expectancy of any product. This is especially true for some applications such as aerospace that must pass rigorous security checking procedures. Geometric reasoning and virtual reality can help in reducing costs and design time by moving testing from physical mock-ups to virtual ones. Originality/value – The paper shows the integration of path-planning techniques in automatic disassembly-planning methods. Keywords Robotics, Simulation Paper type Research paper

physical mock-up. To be useful, these tools should be capable of simulating maintenance operations in a realistic way. The REVIMA project was born with these ideas in mind. Its main goals are the creation of hardware and software tools to realistically simulate maintenance operations on aircraft engines and equipment. The system provides a haptic interface with a big workspace and realistic and interactive visualization of a virtual mock-up (Borro et al., 2004; Savall et al., 2002) (Figure 1). The software provides additional tools that allow certain analyses. For example, a tool was developed to study the allowed workspace at a certain position for the correct operation of a tool such as a wrench (Figure 2). During the simulation of maintenance operations, it is important to establish the conditions under which an operation is possible. For example, the removal of a given part could only be possible after the removal of other components. The REVIMA system allows the user to manually define these constraints. However, this is not an optimal solution, since it is extremely complex to define correctly by hand all the constraints for a long disassembly sequence. An automatic disassembly planner can complement the tools of the REVIMA system by providing additional analysis information, such as the precedence of removals. This information can be exploited in the interactive environment

1. Introduction One important requirement for the aerospace industry is the capability to demonstrate that all their products can be serviced during the operations in an airport. This requirement is translated during the design phase of a product into the necessity of building physical mock-ups of the components. These mock-ups are used to test the maintenance operations for the new designs. Obviously, the cost of these mock-ups is extremely high for products as complex as a turbopropulsor. Furthermore, any defect detected in the physical prototyping step requires returning to the design table to redesign the product and the construction of a new mock-up. The construction of physical mock-ups adds time delays to a project, as well as, an increase in the cost of development of the product. The application of VR techniques can alleviate these added costs by providing an engineering team with tools that allow decision-making prior to the construction of a The current issue and full text archive of this journal is available at www.emeraldinsight.com/0144-5154.htm

Assembly Automation 27/3 (2007) 207– 214 q Emerald Group Publishing Limited [ISSN 0144-5154] [DOI 10.1108/01445150710763222]

207

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

Figure 1 User performing a virtual maintenance task on an airplane engine

planning strategies. The first,one is a simple strategy based on single translations. This procedure is capable of detecting the extraction direction of any component from its contacts, without being limited to a predefined and fixed set of directions. The validation of these directions is accelerated using the graphics hardware of a computer. Contrary to most works in bibliography that require an exact representation of the geometry, this method is capable of dealing with some geometrical errors in the geometrical definition of the mock-up. The second procedure adapts a technique applied in robotics for its use in disassembly planning allowing the generation of complex extraction paths. These methods can be integrated into interactive VR applications for design analysis and validation, training, etc. The rest of this paper is structured as follows. Section 2 of this work presents a small review of some works relevant with the technologies described. Next, we proceed in Section 3 to the detailed description of the proposed automatic disassembly planner. The path planning techniques are described in detail in Section 4. Section 5 presents some results based on the application of the algorithms described to some real world examples, and finally in the conclusions section we review the main ideas derived from this work.

Figure 2 Automatic workspace analysis can mark which areas of a workspace are accessible without collisions

2. Previous work A survey on disassembly planning presented by Lambert (2003), focuses mainly in papers that base the sequencing problem on the product. In this case, the input is not a geometrical representation of the product but the contact graph and precedence information provided by the end-users. The main goal of these works is the analysis of the different sequences to obtain an optimal or near optimal planning, for example by means of optimization techniques such as genetic algorithms (del Valle et al., 2003). Concerning the simulation of assembly or disassembly operations, in Gupta et al. (2001) the authors present a tool to evaluate and simulate assembly plans. This application is divided into four different tools. The assembly editor allows the user to define joints and kinematical restrictions of the assembly. In the plan editor, the user can define an assembly operation in a high level way and the system will automatically ask for any missing information such as specific tool information. The assembly simulator can simulate the operations, performing three main types of operations: interference checking, tool accessibility and path planning for tools. Finally, the results are visualized using a viewer. Automatic assembly planning tools have been successfully integrated with a virtual reality simulation environment such as in the V-Realism program (Li et al., 2003). This software provides a maintenance training tool. The input for the system are the geometry defining files (in STL format), and a graph structure where nodes represent the units to be maintained, and edges store the disassembly constraints for maintenance. This information is used in a disassembly planner to generate optimal assembly plans that the user can inspect using the graphical tools provided. The approaches described above still contain an important interaction with the user for the generation of the disassembly sequences. Other authors propose more automatic methods. For example, in Agrawala et al. (2003) the authors performed a series of psychological experiments to characterize the properties of a good assembly plan from the

to ensure the definition of correct maintenance routines prior to the construction of any physical mock-up. For this application, an automatic disassembly planner generates automatically two types of useful information. First, it generates the paths to remove the components from the product. These paths can be further studied for analysis of accessibility or ergonomics. The second output is the precedence relationship among the different components. These relationships allow the generation of different valid disassembly sequences. The user can study different alternatives using the interactive VR environment to obtain the optimal sequence. Also, the precedence relationships serve to determine when a certain operation is valid. This capability can be exploited in training applications where the automatically generated precedence relationships serve to validate the actions of a trainee. In this paper, we present a new fast disassembly planner capable of using several path planning strategies for the generation of the precedence relationships among the removal of the different components. We propose two path 208

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

point of view of a human. This helped in finding useful design principles that are used by a planner to generate assembly sequences. The system uses a non directional blocking graph (NDBG) structure (Wilson, 1992) to search for subassembly separation directions. The NDBG is a data structure that allows determining how an assembly can be partitioned into two subassemblies. However, this structure is not reliable to floating point errors, or geometric errors contained in the geometry (Latombe, 1999). As we have seen, difficulties dealing with approximate geometry, path planning techniques limited to single translations or incomplete automation are some of the difficulties common in most of the state-of-the-art planners, therefore, limiting their application.

partitioning and detects the contacts among the different components; and a runtime phase that generates the precedence relationships among the removal of the different components and generates the disassembly sequence for the target part. The main goals of the pre-processing phase are the generation of a spatial partition of the scene and the determination of the surfaces in contact among the different parts. The spatial partition also helps in the second goal since it provides support for other algorithms such as the collision detection algorithm used in path planning (Figure 3). Two techniques have been used for the generation of the spatial partition: voxels and octrees. While voxels provide excellent performance for some methods such as collision detection, others, such as the spatial localization of the parts of the assembly benefit from the more structured nature of an octree. The runtime phase includes the proper selective disassembly process. In this phase, the method tries to detect which components can be extracted from the assembly using a trial and error process. The system selects a component from the assembly and tries to remove it. A part can be removed if a disassembly path from the assembled position of the component to the exterior of the assembly can be found. Initially, any part of the assembly can be selected as a potential extraction candidate. However, parts located on the exterior (i.e. a point on the part can be connected with the exterior of the assembly by a path) of the assembly are probably more easy to remove. The system uses the information stored in the spatial partitioning to locate this type of components and it selects them as potential extraction candidates. The path planning problem appears during the determination of whether a part or an assembly can be removed from it. The path planning problem requires the localization of a collision free trajectory that takes the whole component from the assembly and moves outside it. Our selective disassembly system can use two different path planning approaches (Figure 3): a fast single translationsbased path planning or the T-RRT algorithm capable of solving more complex planning problems. These techniques are described with more detail in following sections. This selection and test procedure is repeated until the target component is extracted, therefore, reducing the assembly by removing components layer after layer. This procedure removes more components than required. To avoid non required removals, the system must detect the order dependency, or precedence relationships, among the different removals. A precedence relationship exists between two parts if the removal of one is required for the extraction path of the other to be collision free. With this idea in mind, a simple procedure can be built to detect these relationships. Once an extraction path has been found for a component, the system checks the extraction along that path, taking in consideration all the components of the assembly in their initial configuration. A precedence relationship can be established with all the parts with which a collision is detected when the part moves along the extraction path that was generated when these obstacles were removed. The precedence information, so built, is used to automatically generate the disassembly sequence of the target part.

3. Automatic disassembly planning A selective disassembly planner generates the sequence of part removal required to extract a certain target part from an assembly. This operation is typical in the simulation of maintenance operations, where a part must be removed because it has reached its expected end of life or because it is malfunctioning. To create a selective disassembly planner, several sub problems must be solved. The generation of valid removal sequences requires the generation of collision free extraction trajectories. This problem is called path planning and has been thoroughly studied in robotics (Latombe, 1991). This problem includes the collision detection problem. Performance is one of the most important challenges for both disassembly planning and path planning. The complexity of these algorithms is extremely high, and it can be demonstrated that in their more general definition the related assembly planning problem is NP-complete (Kavraki et al., 1993). The planning of disassembly trajectories is in many planners simplified, and only part removals based on translations along straight lines are allowed (Srinivasan and Gadh, 2002). However, not every assembly contains parts that can be removed along a simple trajectory. On the contrary, some parts require complex removal paths. With these considerations, it is clear that the main challenge for the application of an automatic selective planner in an interactive maintenance simulator comes from the execution time and resource usage. The utility of the system will depend greatly on the capability of the planner to work in a nonobtrusive way. In the following section, we will describe the proposed method. This method can use several path planning strategies, for example the two different planners proposed in Section 4, to generate the precedence relationships and disassembly sequence. 3.1 Disassembly system description and precedence generation The input geometry for the method is defined by a triangle mesh for each part of the assembly, located in its initial assembled position. The method does not require any other input information or user interaction other than the selection of the component targeted for maintenance. The automatic disassembly method consists of two main steps: a pre-processing phase that generates a spatial 209

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

Figure 3 Flowchart of the selective disassembly planner Spatial Partitioning Voxels

Collision Detection

Exterior Objects

Path Planning

Contacts

Octree

Selective Disassembly Planning

T-RRT

Single Translation

Sequence

4. Path planning

evaluates if a given direction meets the conditions provided by equation (1). A function f such the one defined in equation (2) provides an equivalency to the initial problem:

4.1 Single translation path planning A single translation based strategy for path planning assumes that every part can be removed from the assembly along a single straight line. Many assemblies and individual parts from assemblies meet this condition. In this case, the path planning algorithm can be simplified since for one of these trajectories two conditions hold: 1 The local contacts of the part allow motion along the extraction direction. 2 The part does not intersect with other components when it is being removed from the assembly.

f ðxÞ ¼ maxðRð2n i xÞÞ where R(x) is the ramp function defined as: ( 0 if x # 0 RðxÞ ¼ x if x . 0

ð3Þ

For normalised direction vectors x, the values of the function are located in the closed set [0, 1]. The value of the function f is 0 only for extraction directions where all the contacts slide, and therefore, nix ¼ 0, or where the surfaces separate; therefore, the set of solutions of equation (1) is equivalent to the set of values that make the function f evaluate to 0. It is interesting to note that the existence of errors in the definition of the geometry can make this function to always evaluate to values greater than 0. However, in this case, it is usually interesting to evaluate the directions with values of f(x) closer to 0 since it is possible that they are still valid extraction directions, which are being apparently blocked by some contacts. The directions for which f(x) evaluates to 0 or a value close to 0 form the set of potential extraction directions V:

In the first case the condition for a part to have at least one allowed extraction direction is to have at least one solution for the inequation system: ni x $ 0

ð2Þ

ð1Þ

Where ni are the normal vectors of the contact surfaces of the part to be removed and x is a vector representing a potential extraction direction and therefore it is contained in the unit sphere S2. The number and structure of the solutions to this inequation system depends greatly on the set of contacts and a direct approach to solving this system is difficult and prone to numerical errors (Latombe, 1999). Therefore, from a practical point of view, it is useful to transform the problem in equation (1) into an equivalent problem where a function

V ¼ {xj f ðxÞ < 0} 210

ð4Þ

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

The implementation used to find the extraction directions uses a discrete representation of a unit sphere, divided in 18 units in latitude and longitude. The function is evaluated in a matrix of 360 £ 180 points. During the evaluation of these values the system stores in one array the directions of minimal value as a set of potential extraction directions. This implementation is, however, not optimal and it could be improved in the future, for example, using a more uniform sampling (directions near the poles of the sphere are overrepresented in this method), a more intelligent search of solution space that requires less evaluations of the function f or by parallelization the work. Once a set of potential extraction directions V has been found, the method proceeds to validate them testing the complete extraction of the component. This validation is implemented using a procedure based on the projection of the obstacles that uses the graphics hardware Z-Buffer test. A similar approach was proposed in the Archimedes 2 software (Kaufman et al., 1996), but that software did not take into account the relative size of the projections as heuristic to differentiate geometric errors from true intersections, nor used recent techniques such as occlusion queries that helps removing bottlenecks in the transfer of data from the graphics card to the main memory. For an extraction direction to be valid, an observer staying outside the bounding box of the part and looking along the path should not see any other part occluded behind the part being tested. This test can be implemented using the graphics hardware of the computer. The output of this method is the projection of the interference produced by the obstacles as a set of coloured pixels on an output image. If a component can be removed from the assembly without collisions, the image will not contain any marked pixel. If, on the other hand, there is an obstacle component that blocks the removal, the image contains a number of coloured pixels. However, coloured pixels can also appear because of defects in the definitions of the geometry. Usually, in this later case, a simple heuristic criterion can be used to identify them: the number of coloured pixels (the size of the projection of the interference) will be small when compared to the projections (the number of pixels in which the component has been projected) of any of the components. The proportion between the sizes of these projections serves as a heuristic threshold to identify between these possibilities. A basic implementation of the test requires reading the image generated from the graphics card back into the main memory of the computer. This transfer of information is usually slow and, therefore, it can become an important bottleneck. The usage of a technique called occlusion queries available in recent hardware can help in the elimination of this bottleneck. As the name implies, occlusion queries are used mainly for determining if a part is visible or occluded. With this purpose the program sends a simple approximate shape of the space occupied by that component to the graphics card and the graphics card calculates if any of the pixels of that shape are visible under current conditions. These queries return the number of pixels that have been rendered. We can adapt this method to our problem to size the projections of the different components and of the interference without reading and processing an image. Since, the method works in a picture space and not in the original 3D space, the detection of interferences depends on

the resolution of the image used. For example, a thin component, like a ring, can have a 0 pixel projection, and therefore its removal cannot be correctly predicted. In practice, with output images of 512 £ 512 or bigger, we have not encountered this type of problem. 4.2 Complex path planning As stated above, not every disassembly problem can be solved by simply translating a part along a straight trajectory. In many assemblies, and from a general point of view, the extraction of a component from an assembly requires following a complex trajectory. Several techniques have been proposed to solve the path planning problem (Latombe, 1991). Some of the most successful algorithms are based on a stochastic search of the configuration space of the mobile object, as the rapid growing random trees (RRT) algorithm (Kuffner and LaValle, 2000). The initial RRT algorithm generates two random trees that grow from the initial and final configurations. In every step a new random configuration is generated. One tree is extended towards this configuration from its nearest point. If the connection is possible the new point is added or it is discarded otherwise. The system checks if the two trees can be connected to generate the solution path. Speed constraints can be added to this basic method to generate the motion of robots with kinematical constraints (LaValle and Kuffner, 2001). Also parallel implementations of the method have been presented (Carpin and Pagello, 2002). RRT type algorithms do not require pre-processing, and usually the memory consumption is low, since only the tree structure must be generated and maintained. Furthermore, the algorithm is easily interruptible and parallelizable. However, the algorithm cannot guarantee the location of a solution path within a finite time, and the generated paths are not guaranteed to remain constant across executions. Also, the execution time of the algorithm depends on the specific execution. In addition, the resulting paths are not optimal. However, this limitation is not important for the disassembly planning applications, since, in this case, it is usually more important to obtain a valid extraction path that allows the generation of new sequences than having a smooth path. The approach used in this work for path planning is based on a modification of RRT called extended-RRT (ERRT) (Bruce and Veloso, 2002) algorithm, used online in robots for motion planning. As in ERRT, our algorithm only uses a single expansion tree. The method selects with a probability 1-p random points within the bounding volume of the scene. The nearest point in the tree to these random configurations is located. To avoid long jumps the tree is not extended to the targets but to points at a random distance from the tree on the connection line. If the new position is valid (i.e. there are no collisions) the new configuration is linked to the tree. However, the main difference with ERRT comes from the fact that our problem does not contain a defined final target configuration for the path. In the case of ERRT this target configuration is selected with a probability p, this guides the search towards the goal. In our case, however, since we do not have a defined target configuration for the disassembly path planning problem, the tree is extended towards points exterior to the axis aligned bounding box (AABB) of the assembly. 211

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

The method with these modifications is called T-RRT or target-less RRT, since the method does not require a defined target configuration. Figure 4 represents an example where a gear has been removed from its casing. The figure displays the generated search tree.

The execution time depends on the complexity of the model (i.e. number of polygons and parts) but also, it is interesting to note that the execution times for two problems of the same geometrical complexity such as torque converter and admission are quite different. In these cases, the difference comes from the different structure of the models. The admission model contains many components located on the exterior that are readily accessible and removable. These components are selected and removed during the process before the precedence relationships are completed and therefore the final sequence can be established. The torque converter on the other hand contains few potentially removable parts and therefore, the disassembly is faster. For most problems the execution time is low and therefore the system can be integrated without problems in an interactive environment. The system would have a small delay but it would remain responsive. Even in the most complex examples, the execution time is less than 100s. Even if this would interrupt the normal execution of the interactive environment, the added information provided by the planner allows more complex testing environments. Figure 6 displays a typical maintenance operation, with the disassembly steps required to service the air filter of an engine. The procedure starts by removing of two keys (not shown in this figure but visible in Figure 7). Once they have been removed a fixture can be disassembled from the assembly (first 3 screenshots). Afterwards the cover protecting the air filter can be removed. This cover allows the access to the air filter elements which is removed in the last steps. The models used for testing T-RRT can be seen in Figures 4 and 7. Table II resumes the average, maximum and minimum execution time found using this algorithm. Even though the average query time is not extremely high, the stochastic nature of the algorithm can generate some cases of extremely long execution. This execution time depends again on the geometrical complexity of the part being removed, but also on the structure of the extraction path. For example, parts whose extraction path needs to cross a narrows section such as gear are more complex and therefore the execution time is longer. The execution times of Table II only include the removal of a single part. Therefore, it is clear that the application of a general path planner like the T-RRT for every component of a product would produce extremely long execution times. In this case it seams more reasonable to apply a complex path planning for those parts that are previously known to have a complex extraction path, while using a simpler strategy for the general disassembly planning problem.

5. Results All the tests described in this section have been performed on 3 GHz Intel Pentium 4 machines with at least 1GB of RAM and graphic cards based on the nVidia GeForce 6800 chipset. All the development and test has been performed under Microsoft Windows XP. Table I displays the execution time of the proposed method using a simple translation strategy for the removal of every component. The different models are represented in Figure 5. Figure 4 Initial configuration and the search tree generated by the T-RRT for the extraction of a gear from its casing

Table I Number of components in the assembly, number of elements removed, number of triangles of the model and the execution time in seconds of the disassembly planner using a single translation strategy Model Box Valve Torque converter Admission Differential

Parts

Removed

Triangles

Time (s)

9 11 36 69 78

8 9 5 6 20

3,003 12,618 224,887 190,809 150,171

2.26 1.82 2.68 35 98

6. Conclusions We present a fast selective disassembly planner that can be integrated into a maintenance simulation tool to provide additional analysis information during the design and validation of a product. The method generates the precedence information required to generate valid disassembly sequence, and therefore, allows several applications such as generation and analysis of potential sequences generated automatically, optimization of 212

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

Figure 5 From left to right different models used in testing: box, valve, torque converter, admission and differential

Figure 6 Disassembly of the air filter of the admission example

Figure 7 Removal path of a key fixing the air filter of an engine

Table II Average, maximum and minimum execution times in seconds for two examples of the T-RRT algorithm Model Key Gear

Average (s)

Max. (s)

Min. (s)

3.7 14.7

37.6 538

1.5 3.97

the disassembly sequence or the analysis of different type of problems in the disassembly process. The method proposed is capable of generating a disassembly sequence for models of medium complexity in a completely automatic way in less than 2 min. More complex examples would benefit from the interaction of the 213

Path-planning techniques for the simulation of disassembly tasks

Assembly Automation

Iker Aguinaga, Diego Borro and Luis Matey

Volume 27 · Number 3 · 2007 · 207 –214

engineer that could reduce the number of potential components by selecting components or areas of interest. We have adapted a method from robotics for its use in disassembly planning and created the T-RRT algorithm to solve complex disassembly path planning problems. However, the usage of complex path planning should still be reserved for parts whose path is known to be complex, since the added complexity of this method extends the processing time.

Kavraki, L.E., Latombe, J-C. and Wilson, R.H. (1993), “On the complexity of assembly partitioning”, Information Processing Letters, Vol. 48 No. 5, pp. 229-35. Kuffner, J. and LaValle, S.M. (2000), “RRT-connect: an efficient approach to single-query path planning”, Proceedings of the IEEE International Conference on Robotics and Automation 2000 (ICRA 2000), pp. 995-1001. LaValle, S.M. and Kuffner, J.J. (2001), “Randomized kinodynamic planning”, International Journal of Robotics Research, Vol. 20 No. 5, pp. 378-400. Lambert, A.J.D. (2003), “Disassembly sequencing: a survey”, International Journal of Production Research, Vol. 41 No. 16, pp. 3721-59. Latombe, J-C. (1991), Robot Motion Planning, Kluwer Academic Publishers, Boston, MA. Latombe, J-C. (1999), “Motion planning: a journey of robots, molecules, digital actors, and other artifacts”, International Journal of Robotics Research, Vol. 18 No. 11, pp. 1119-28. Li, J.R., Khoo, L.P. and Tor, S.B. (2003), “Desktop virtual reality for maintenance training: an object oriented prototype system (V-REALISM)”, Computers in Industry, No. 52, pp. 109-25. Savall, J. et al. (2002), “Description of a haptic system for virtual maintainability in aeronautics”, Proceedings of the 2002 IEEE/RSJ International Conference on Intelligent Robots and Systems, Lausanne, Switzerland, pp. 2887-92. Srinivasan, H. and Gadh, R. (2002), “A non-interfering selective disassembly sequence for components with geometric constraints”, IIE Transactions, No. 34, pp. 349-61. Wilson, R.H. (1992), “On geometric assembly planning”, PhD thesis, Stanford University, Palo Alto, CA.

References Agrawala, M. et al. (2003), “Designing effective step-by-step assembly instructions”, Proceedings of the Computer Graphics, SIGGRAPH, San Diego, pp. 828-37. Borro, D. et al. (2004), “A large haptic device for aircraft engines maintainability”, Vol. 24, pp. 70-4. Bruce, J. and Veloso, M. (2002), “Real-time randomized path planning for robot navigation”, Proceedings of the IROS 2002, Switzerland. Carpin, S. and Pagello, E. (2002), “On parallel RRTs for multi-robot systems”, Proceedings of the 8th Conference of the Italian Association for Artificial Intelligence, pp. 834-41. del Valle, C. et al. (2003), “A genetic algorithm for assembly sequence planning”, in Mira, J. (Ed.), Lecture Notes on Computer Science 2687, Springer, Berlin. Gupta, S.K., Paredis, C.J.J. and Sinha, R. (2001), “Intelligent assembly modeling and simulation”, Assembly Automation, Vol. 21 No. 3, pp. 215-35. Kaufman, S.G. et al. (1996), “The Archimedes 2 mechanical assembly planning system”, Proceedings of the IEEE International Conference on Robotics and Automation 1996 (ICRA 1996), Minneapolis, Minnesota (USA), pp. 3361-8.

To purchase reprints of this article please e-mail: [email protected] Or visit our web site for further details: www.emeraldinsight.com/reprints

214

Path-planning techniques for the simulation of ...

Its main goals are the creation of hardware and software tools ... visualization of a virtual mock-up (Borro et al., 2004; Savall .... characterize the properties of a good assembly plan from the. Figure 1 ... The NDBG is a data structure that allows.

301KB Sizes 1 Downloads 200 Views

Recommend Documents

Path-planning techniques for the simulation of ...
directions. The validation of these directions is accelerated using the graphics hardware of a computer. Contrary to most ... as in the V-Realism program (Li et al., 2003). This software ..... (Bruce and Veloso, 2002) algorithm, used online in robots

Parallel generation of samples for simulation techniques ... - CiteSeerX
The current SAN solver, PEPS software tool [4], works with less than 65 million ..... multiprocessor technology machine and perform the three simulation approaches for a ..... ular Analytical Performance Models for Ad Hoc Wireless. Networks.

Distributed Modelling Techniques for System Simulation
The introduction of software tools for model specification has greatly ..... would be marginally more computationally demanding than an analytical solution.

Simulation of Handoff Techniques in Mobile Cellular.pdf ...
average signal pattern due to scatters and different terrain variations as the mobile. moves. The slower fading effect (shadow fading) follow log-normal distribution. Hence handoff methods in the presence of multipath and shadow fading usually. resul

Monte Carlo Simulation for the Structure of Polyolefins ...
unsaturated (usually vinyl) groups can be incorporated by the. CGC into a ... broaden the molecular weight distribution, since chains formed at the second site will .... to published experimental data for a lab-scale synthesis of a dual catalyst ...

development of a dynamic simulation tool for the ...
Space Agency (ESA) ExoMars mission, will require rovers to travel further, faster, ... on-board software, or operator-in-the-loop simulations. Although ROAMS ...

The discrete dipole approximation for simulation of light ...
The code is written for distributed memory systems using the message passing interface (MPI).2 ... is incident electric field, ¯Gij is the free-space Green's tensor. (complex symmetric), and .... to the ''rule of thumb'' (see main text for explanati

Zero Mach Number Diphasic Equations for the Simulation of Water ...
for the Simulation of Water-Vapor High Pressure Flows. Stéphane ... Montréal C.P. 6128, Succ. Centre-Ville, Montréal QC, H3C 3J7 ... We call the resulting set of ...

development of a dynamic simulation tool for the ...
problem and discuss the development of a planetary rover chassis simulation tool ... on-board software, or operator-in-the-loop simulations. Although ROAMS ...

Toward a simulation-based tool for the treatment of ... - ScienceOpen
May 2, 2011 - the most common illnesses that affects voice production is partial weakness of the ..... of the IBM can be found in Reference (Mittal et al., 2008). To minimize the use ...... access article subject to a non-exclusive license between ..

reservoir simulation mathematical techniques in oil recovery 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. reservoir ...

Comparative evaluation of drying techniques for surface
Universiw of California, Los Angeles, CA 90095-1597, USA. Abstract. Five different ... 'stiction' in the field of microelectromechanical systems. (MEMS). Stiction ...

Comparative evaluation of drying techniques for surface
testing of a C-shape actuator, Tech. ... John Y&in Kim was born in 197 I. He received his B.S. in ... B.Tech. degree in mechanical engineering from the Indian.

A Study Of Various Techniques For The Brain Tumor ... - IJRIT
IJRIT International Journal of Research in Information Technology, Volume 3 ..... assignment and there is a need and wide degree for future examination to ... Journal of Advanced Research in Computer Science and Software Engineering, Vol.

On the Development of Name Search Techniques for Arabic.pdf ...
Such systems are used for applica- tions as diverse as tax fraud detection and immigration con- trol. Using names to retrieve information makes such systems.

Call For Papers: Techniques of the Corporation Conference
We invite emerging and established scholars in diverse fields (including business history; labour history; anthropology; geography; economic sociology; media studies; critical race studies; architecture studies; feminist and sexuality studies; enviro

Fusion of micro-metrology techniques for the flexible ...
Cell phones, for instance, are ... most important “non-contact” micro-metrology techniques (optical and non-optical), performing a comparison of these methods ...

Techniques for Improving the Performance of Naive ...
... negatively. In such cases,. 1 http://people.csail.mit.edu/people/jrennie/20Newsgroups/. 2 http://www.cs.cmu.edu/afs/cs.cmu.edu/project/theo-20/www/data/ ...

Techniques for Improving the Performance of Naive Bayes ... - CiteSeerX
and student of the WebKB corpus and remove all HTML markup. All non-alphanumeric ..... C.M., Frey, B.J., eds.: AI & Statistics 2003: Proceedings of the Ninth.