Scorpius Team Description Paper Virtual Robot Competitions, China 2008 Khashayar Niki Maleki1, M. Hadi Valipour1, Roohollah Yeylaghi Ashrafi1, Sadegh Mokari1, S. Ali Zahiri1 and M. R. Jamali2 1

Department of Electrical and Computer Engineering, Faculity of Engineering, Shahid Rajaee University, Tehran, Iran 2 Control and Intelligent Processing Center of Excellence, Department of Electrical and Computer Engineering, University of Tehran, Tehran, Iran {kh.niki, m.h.valipour, r.yeylaghi, s.mokari, sa.zahiri}@sru.ac.ir, {m.jamali}@ece.ut.ac.ir

Abstract. This paper is aimed to summarize some implementation processes and main features of Scorpius simulated virtual agents, including detailed specifications and algorithms. This is the third year that Scorpius team is participating in Robocup simulation competitions, but it is our first1 experiment in Virtual Robot Simulation League. For this competition we tried to implement a simple team, where our researches are mostly focused on exploration, mapping, removing noise effects from the map, victim localization and applying different kinds of robots for these purposes.

Keywords: SLAM, virtual robot simulation, Kalman filter, occupancy grids, topologically organized map, localization, mapping, hybrid

1 Introduction Every year in all around the world, thousands of people die or get injured under collapsed buildings or other unreachable areas because of explosion, earthquake, flood and other natural or unnatural disasters. Delay in Rescue operations due to dangerous situations, vast locations, massive casualties, etc. leads us to employ high technologies to aid rescue group. Designing, implementing and using rescue robots not only makes explorations much more accurate by seeking for vital signs of victims, but also brings much more safety to rescue teams. The first Robocup Rescue project was initialized after the Great Hanshi-Awaji earthquake which hit Kobe City (Japan) on the 17th of January 1995 and caused more than 6500 casualties. Since 2002, RoboCup Rescue Competitions are held as part of the annual RoboCup World Championships. It is the purpose of RoboCup Rescue, to promote research and development in this socially significant domain in order to 1

We also intend to participate in Iran Open 08 competitions before China 2008

2 Khashayar Niki Maleki1, M. Hadi Valipour1, Roohollah Yeylaghi Ashrafi1, Sadegh Mokari1, S. Ali Zahiri1 and M. R. Jamali2

ultimately acquire solutions that can be used by USAR (Urban Search and Rescue) teams under real emergency circumstances. Current Scorpius virtual robot team uses P2DX and Talon robots equipped by SONARs, SickLMS, Odometry, INS, RFID, Victim Sensors, etc. for their explorations, mapping and victim detections. Now we are working on new Air Robot for mapping, planning and even victim detection by image processing. In this paper at first, the way our robots and server communicate and its basic architecture is denoted, then SLAM (simultaneous localization and mapping) and the main approaches which are applied to design map from the unknown environment, next section discusses victim detection, where section 5 and 6 talk about Applicability to real robots and reusability and finally chapter 7 concludes the paper and describes future works.

2 Communication In virtual robots just like other simulation environments there are two main parts; the agent and the server. The Agent represents one or more robots with sensors and actuators which get sense from and affect the virtual environment which simulator (server). Fig.1 describes the way server and robots communicate with each other.

A

B

Fig. 1. The architecture of server side (A) and client side (B) of connection

This decentralized setup allows us to spread the computations across many machines and still run a relatively large team of agents. Only one server instance is required, which will record the pre-processed mapping information from the agents. Robots employ wireless communication in order to share some information about the region they are exploring. Although this communication is limited by distance, but sometimes it helps robots to minimize overlapping in explored areas and improves coordination in multi-agent fields.

Scorpius Team Description Paper

Virtual Robot Competitions, China 2008 3

3 Localization, Mapping, and Exploration Intelligent robot needs to use SLAM techniques to explore and rescue in real or virtual environments. The first question is Where am I? answer to this question solves localization and the other question is How is my environment? which suggests mapping. So SLAM is a technique that is employed by robots to design a map and create a world model from the unknown environment. SLAM implementation includes several steps. The data which robot receives from its sensors are fully noisy and it defects the performance considerably. So first of all we should make the data smooth and eliminate noise from it. We proposed Kalman filter as a good solution for it [1, 2, 3, 14]. On the other hand decomposition of the map in to several sub-maps is another approach; this method called occupancy grid, finds landmarks in the main map and then creates grids on it.

X

X

t

Yt

Yt

t

1

MA

1

(1)

t

(2)

Y actuator

In (1) and (2) the map has updated due to the linear motion; At is to store the current motion commands. Yactuator is to denoted the associated uncertainly . Zt : Current looking Kt : gather the estimate distance

K

YtC t (C YtC

t

X

X

t

Yt

t

K t(Z

t

(I

Y measure ) C X

t

K tC

t

)Y

t

t

)

1

(3) (4) (5)

In (3,4,5) Xt , Yt is the result that get from Kalman filter and Xt-1 , Yt-1 is the input data for Kalman filter. Occupancy Grids has some benefits; The most important one is that grid-based representations of a robot s physical environment can be used directly by most navigation, obstacle-avoidance and learning algorithms. Another key benefit of occupancy grids is that the resolution can be tuned to represent the environment s geometric properties at any desired amount of detail. In theory, this is bounded above only by the level of detail originally captured by the sensors. This property of occupancy grids makes them the ideal metric representation for maps that should

4 Khashayar Niki Maleki1, M. Hadi Valipour1, Roohollah Yeylaghi Ashrafi1, Sadegh Mokari1, S. Ali Zahiri1 and M. R. Jamali2

contain a high amount of detail, a feature that most other map representations lack [11, 12]. To accomplish occupancy grids, there are sensors in the robot that use a simple technique with ray-casting .This beam has been sent directly from the robot to obstacle to determine distance between robot and obstacle. For example, if an obstacle is detected at some relative distance, then rays can be cast from the cell at the current position of the robot towards every cell that coincides with the detected obstacle. For all the cells which intersect with the cast rays before the rays hit on the obstacle, the miss counter is incremented. Likewise, for the cells that intersect with the detected obstacle the hits counter is incremented [13]. Afterwards, the occupancy of every grid cell can be determined by threshold of the ratio of the number of hits over the sum of hits and misses together (6).

Occupancy of each cell

Hits

Hits Misses

(6)

After occupancy grids, another approach for localization will be employed, which is based on Topologically Organized Map. In this step a special graph will be created from the path of robot, In addition knowing that the obstacle is detected, a diagram will be created from the map to divide it in to the sub-maps. These sub-maps have no overlap with each other and each one describes a safe area for robot to move in that region [6, 8]. The first advantage of using topologically map is its outstanding support for path planning algorithms. Autonomous robots need to go from a place to another in order to complete their designated tasks. Considering the task of exploration which all robots need to perform, it is conceivable that in some points the robot reaches a deadend. In this situation before it can continue to search another area, the robot has to get out of that location first. Topological maps facilitate these kinds of tasks very well. The nodes and links in graphs and diagrams tell any path planning algorithm precisely how to go through the explored areas safely. The second advantage is their compactness. They are capable of representing huge environments in a very compact way. While the size of occupancy grids and featurebased maps grow exponentially with the size of the environment or the number of detected features, topological maps typically only grow linearly in size as nodes and links are added to denote recently explored areas.

Fig. 2. Applying four steps for localization and mapping [13]

Scorpius Team Description Paper

Virtual Robot Competitions, China 2008 5

So as Fig.2 shows, we can summarize the steps of mapping and localization; Applying Kalman filter on feature based map. Making a grid cells from the map. Implement topological map to determine obstacles and safe paths. Decompose the main map and start mapping from unknown environment and design a new map. So, each robot performs exploration, mapping and localization with the algorithm described above and creates its own sub-map. Finally, all of the maps which are separately designed merge with each other to achieve the main map (7). Main Map = Merge (Sub-map1 , Sub-map2 ,

.. , Sub-mapn )

(7)

Using topological map we have required information for planning, there are some methodologies proposed for planning regarding; A* algorithm [15] and other AI approaches, heuristics, classic algorithms based on graphs i.e. Shortest Path Algorithms (e.g. Dijkstra, Floyd), Graph Traverse Algorithms (e.g. BFS, DFS, etc.) In our limited time we found our heuristic algorithm much more efficient both in functionality and amount of processes.

4 Victim detection One of the important sections in scoring every team is finding victims, so we employed different types of sensors for victim detection (i.e. Sound, Touch, Human motion and Victim-and-False-positive sensors). Our algorithm is implemented considering this probability that using only victim-and-false-position sensor or touch sensor may lead to false detection or latency in finding victims. Every victim may have motion or make sound, so it is possible for Human motion or sound sensor to help us finding victims. One of the benefits of using sound sensor in virtual environment is that, there is only one source for sound generation which is victims [13], so it can help each robot to detect victims with more accuracy .

5 Applicability to Real robots In real world we are facing some problems which are usually ignored or not seriously considered in simulation, regarding: rather low or limited power of processing, fully noisy fields, limitation in employing different types of sensors, costs, etc. We tried to minimize CPU consumption by optimizing our SLAM algorithm besides using well implemented Kalman filter which helps us reduce effects of noise and decreasing number of employed sensors and cameras with not losing considerable accuracy but saving power and process.Utilizing different kinds of robots also helps us to find out

6 Khashayar Niki Maleki1, M. Hadi Valipour1, Roohollah Yeylaghi Ashrafi1, Sadegh Mokari1, S. Ali Zahiri1 and M. R. Jamali2

which robot can perform better in different situations and this can improve efficiency in real world with much less costs.

6 Reusability We developed our project on C# and created so many components in each layer of team structure, these components can easily be utilized in other projects and also can be modified whenever it s required, without need to change any other parts. This component based-layer architecture increases reusability of our code. We also tried to have reusable robot configurations, where changing the configuration of robot for each map may harm the performance or lead to logical bugs like making mistakes in measurements, etc.

7 Conclusion and Future works In this paper we demonstrated an overview of the Scorpius virtual simulation agent design. Most of the efforts were focused on mapping, noise elimination and localization in rescue simulation Virtual environment. First, communication and some basics were introduced then some applicable approaches and useful algorithms to discover the unknown world around were covered. However, Scorpius agent still has a long way ahead to approach the desired behavior. We hope to implement fully autonomous agents with distributed decision making system which can give a proper map in less time and more accuracy by developing our SLAM algorithm and employing heterogeneous robots and more reliable victim detection skills.

References 1. J.A. Castellanos and J.D. Tardos. Mobile Robot Localization and Map Building: A Multisensor Fusion Approach. Kluwer Academic Publishers, Boston, MA, 2000. 2. G. Dissanayake, P. Newman, S. Clark, H.F. Durrant-Whyte, and M. Csorba. A solution to the simultaneous localisation and map building (slam) problem. IEEE Transactions of Robotics and Automation, 2001. 3. H. Durrant-Whyte, S. Majumder, S. Thrun, M. de Battista, and S. Scheding. A Bayesian algorithm for simultaneous localization and map building. In Proceedings of the 10th International Symposium of Robotics Research (ISRR01), 2001. 4. Giorgio Grisetti, Cyrill Stachniss, and Wolfram Burgard. Improving grid-based slam with rao-blackwellized particle filters by adaptive proposals and selective resampling. In Proceedings of the IEEE International Conference on Robotics and Automation, 2005. 5. D. H¨ahnel, D. Fox, W. Burgard, and S. Thrun. A highly efficient fastslam algorithm for generating cyclic maps of large-scale environments from raw laser range measurements. In Proceedings of the Conference on Intelligent Robots and Systems (IROS), 2003.

Scorpius Team Description Paper

Virtual Robot Competitions, China 2008 7

6. Andrew Howard, Lynne E. Parker, and Gaurav S. Sukhatme. Experiments with large heterogeneous mobile robot team: Exploration, mapping, dployment and detection. International Journal of Robotics Research, 25(5):431 447, May 2006. 7. R. E. Kalman. A new approach to linear filtering and prediction problems. Trans. ASME, Journal of Basic Engineering, 82:35 45, 1960. 8. B. Lisien, D. Morales, D. Silver, G. Kantor, and H. Rekleitis, I. andChoset. The hierarchical atlas. IEEE Transactions on Robotics and Automation, 21:473 481, 2005. 9. M. Montemerlo, S. Thrun, D. Koller, and B. Wegbreit. Fastslam 2.0: An improved particle filtering algorithm for simultaneous localization and mapping that provably converges. In Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence (IJCAI), Acapulco, Mexico, 2003. IJCAI. 10. J. Nieto, J. Guivant, E. Nebot, and S. Thrun. Real time data association for fastslam. In Proceedings of the IEEE International Conference on Robotics and Automation, 2003. 11. Sebastian Thrun and Michael Montemerlo. The graph slam algorithm with applications to large-scale mapping of urban structures. Int. J. Rob. Res., 25(5-6):403 429, 2006.ISSN 0278-3649. 12. Sebastian Thrun, Wolfram Burgard, and Dieter Fox. Probabilistic Robotics (Intelligent Robotics and Autonomous Agents). The MIT Press, September 2005. ISBN 0262201623. 13. ManifoldSLAM: a Multi-Agent Simultaneous Localization and Mapping System for the RoboCup Rescue Virtual Robots Competitionwritten by Bayu Slamet Max Pfingsthorn Master in Artificial Intelligence at the Universiteit van Amsterdam.Date of the public defense: December 11, 2006. 14. S.Williams, G. Dissanayake, and H.F. Durrant-Whyte. Towards terrain-aided navigation for underwater robotics. Advanced Robotics, 15(5), 2001. 15. StuartRussell, PeterNorvig, ArtificialIntelligence:AModernApproach, Pearson Education, (2002.4).

Scorpius Virtual Robot Team Description Paper

The architecture of server side (A) and client side (B) of connection ... Robots employ wireless communication in order to share some information about ... Occupancy Grids has some benefits; The most important one is that grid-based.

97KB Sizes 0 Downloads 244 Views

Recommend Documents

Scorpius Virtual Robot Team Description Paper
dangerous situations, vast locations, massive casualties, etc. leads us to employ .... of the IEEE International Conference on Robotics and Automation, 2005. 5.

Scorpius Virtual Robot Team Description Paper
Current Scorpius virtual robot team uses P2DX and Talon robots equipped by .... Proceedings of the IEEE International Conference on Robotics and Automation, ...

Scorpius Team Description Paper Soccer Simulation ...
approach, genetic algorithm, fitness function, data mining, association rule. ... methods using Data Mining techniques which is presented in 2.3. 2.1 Trajectory Based .... Han, J., Kamber, M.: Data Mining Concepts and Techniques. 2nd edn.

Scorpius Team Description Paper Soccer Simulation 3D ... - CiteSeerX
ing, tackling, dribbling, and ultimately, since the whole game is a team work thus team coordination and the issues concerning multi agent decision making ... The rhythmic component of the gait is described by a coupled oscillators system, modeling t

Scorpius Team Description Paper Soccer Simulation ...
shape, having the same locomotion and even more capabilities has always been one of the ... methods using Data Mining techniques which is presented in 2.3.

Scorpius 3D Team Description Proposal
The recent RoboCup simulation environment named “spark” is a good platform for implementation of robots and the methods employed for controlling them.

Team Description Paper SBCe_Saviour Virtual Robots ...
system and the operator graphical interface respectively. ..... multiple robots' mapping data, sensor data, camera images, battery status and ... Cisco Press., 1998.

ENBank Team Description Paper
The recent RoboCup simulation environment named spark is a good .... The first formula we need is of course the formula of the position of the CoM for a solid:.

Zigorat 2007 Development Team Description
This paper describes the main contributions of the Zigorat. Soccer 3D development team that is going to take part in ... paper describes our studies about achieving a stable walking pattern for legged-sphere agents as well as our works with .... fina

Zigorat 2007 Development Team Description
We plan to use Webots as a simulator for biped modeling. The mobile robotics simulation software named Webots enables convenient and rapid modeling, pro- gramming and simulating various types of mobile robots[2]. In the same paper, the author (and pr

Zigorat 2007 Development Team Description - Chaosscripting
Soccer 3D development team that is going to take part in RoboCup competitions this year. ... attempt to speed up the development of the current simulation environment by encouraging members of .... namics, Stability, Control and Application.

HELIOS2007 Team Description - WrightEagle 2D Soccer Simulation ...
tool. Our positioning mechanism has many advantages than previous methods. 1 Introduction .... training process and enables us to edit the training data easily.

HELIOS2008 Team Description - WrightEagle 2D Soccer Simulation ...
and can be adjusted by human's intuitive operations using GUI tools. .... BI| = n2. 3.3 Feature of Basic Model. The basic model can perform with light-weight ...

Zigorat 2007 Development Team Description - Chaosscripting
Soccer 3D development team that is going to take part in RoboCup .... design; the developer has to decide on the type (biped, quadruped etc.), size, shape ... Besides the cost aspects compared .... namics, Stability, Control and Application.

WrightEagle2008 Simulation 2D Team Description ...
We take simulation 2D as a typical problem of multi-agent systems, and our ... of decision-making at the end of the cycle, so he will do nothing in the current .... considered as a potentially good forward and given a high evaluation value. In.

Team Blazers Paper
Nov 24, 2013 - Submitted to Colorado Technical University On… Student Paper. Submitted to American Public University System. Student Paper www.tgiltd.

Robocup Austria 2009–Rescue Simulation League Virtual Robot ...
ETB Robotic Association Labs, Computer Engineering Department ... Because of mentioned problems, shortage of equipments, cost overheads and also.

Robot predators in virtual ecologies: the importance of ...
Robot predators in virtual ecologies: the importance of memory in mimicry studies .... few data about the relationship of stimulus intensity and forgetting rates, we ...

RoboCup Rescue 2008 - Robot League Team MRL ...
while it is stable. So we've developed 4 remote controlled robots NAJI-I, NAJI-II, .... In Proc. of the IEEE Computer Society Conference on Computer Vision and.