Distributed Decision-making Control System For The Cooperative Task Of Multi-mobile Robots

Cooperation is one of the most important topics of multi-robot. In the thesis, the problems such as system architecture, the cooperation and coordination in task layer, the way of planning in behavior layer are discussed based on the 'pursuit-evasion' game.The thesis launches from task layer planning and behavior layer planning. A novel algorithm is presented for task decomposition and assignment, which is used in pursuit and sending task. The argument of its superiority and the analysis of its complexity are given as well. Based on the 'constriction algorithm' used in searching for the target in the image, the 'stretch and constriction algorithm' is presented for target search, which receives a better result. The control policy of the behaviors such as avoiding and pursuing is the combination of potential field method and behavior fusion method. The reinforcement learning based on existing knowledge is used to modify the importance parameters of each behavior, which will avoid the disadvantage of rules.Finally, a simulation system is built to realize the multi-robot system. And the feasibility of all algorithms is verified further.