With the development of robotics, the power of robot has been improved and the application areas of robot have been enlarged. Robots are required to accomplish more complex tasks, such as assembly, space station restore, ocean exploration and toxic waste cleanup. To accomplish those tasks, a system that consists of multiple simple robots has more advantages than a single powerful robot. In order to work efficiently, that system should be organized. Therefore, how to realize the coordination and cooper~ation of multiple robots is one of the new issues of robotics. This paper is focused on multi-robot motion planning and the implementation of a genetic simulation platform. The content of this paper goes as follows: Firstly, the features of multi-robot system are described, researches on motion planning and multi-robot simulation are reviewed, and the architecture and background of this thesis are introduced. Secondly, distributed motion planning based on behavior is addressed. Artificial Potential Field approach is applied to the robot to avoid collision with obstacles. By (,living different weight to the obstacles in different directions, the path planned by the robot is improved. To solve the problem of minimum in environment, follow_wall behavior is introduced, and by introducing Goal Angle, the robot can easily get out of minimum in more complex environment. Collision avoidance between robots is also discussed in this paper. By exploiting reasonable traffic rules, robots are able to avoid each other agilely. Thirdly, Learning Classifier System (LCS) is applied to multi-robot systern. To improve the speed of LCS, Rule Constructor and Merge operation are introduced. In addition, robots shared the optimal rules by communication among them, by which the parallel features of both multi-robot system and LCS are exploited. The simulation results show that the speed of LCS is improved. Fourthly, the design and implementation of multi-robot simulation system are studied. The object model, dynamic model and function model of simulation system are built by exploiting Object Modeling Technique (OMT) and a distributed genetic multi-robot simulation platform is developed on the base of those models. Finally, a conclusion is given and future work is addressed.
修改评论