Study On Multi-robots Path Planning Based On Agent Theory

As science and technology progresses, the robot technology is developing at a high speed. Because of the restriction in capability and space for single robot, multi-robots come in to being. Research shows that it's more economic to develop simple multi-robots than single complex robot to solve some dynamic and difficult problems. And besides, with the birth of robot product line and the need of manufacture flexibility, people are eager to achieve robot's self-determination.Path planning is the precondition of accomplishing missions logically and efficiently for multi-robots. How to choose reasonable or even optimal path is important question for study. Multi-robots path planning works in the multiple robots system. Its role is to program paths for every robot while ensuring no collision between robot and the environment or between robots all the time. This paper will study path planning method in multi-robots system based on Agent theory.Firstly, this paper introduces the multi-robots system, the multi-agents system (MAS) and the development and characteristic of the multi-robots system. Then it introduces the path planning methods of multi-robots, emphasizing on some path planning methods, which are the focus of researchers. After this, we propose a path planning method called Improved Artificial Potential Field (IAPF) method. To solve the "collision avoidance lock" for multi-robots, we add a rotating force to the robot on the base of attracting force and repulsing force in traditional APR When the robots are too near, the three forces will make all of robots in danger turn right (or left). So this avoids their being near again. Afterwards, we carry out a simulation to test the method and prove its validity. We also find out that the path lengths are a little bit long. To improve it, we propose another method based on Genetic Algorithm and IAPF, which we call mixing method. We regard every robot as an Agent, which can distinguish the type of bars. Then we use GA to plan routes for all robots respective on the whole. When some robots are too near, we use IAPF to negotiate them locally. We carry out a simulation and find the mixing method is better than IAPF. Finally, weplan the future work to improve the both methods.