Design Of Multi-robot Collaboration In Robot Soccer Environment

With the increasing application of multi-robot system, more and more people were concerned with multi-robot cooperation. As a key part of multi-robot system, multi-robot cooperation had a direct effect on the efficiency, the cost and the performance of running the system. As a multi-agent system, soccer robot system was a challenging research platform, which combined with many technologies, such as real-time vision manipulation, wireless communication, multi-robot cooperation, etc. Multi-robot cooperation in the robot soccer environment could be applied to many fields, and it had practical significance.In this thesis, multi-robot cooperation in robot soccer environment of the Middle League MiroSot games was mainly researched, which was to coordinate behaviors of many individual robots to accomplish the common mission of defense and goal by the rules of robot soccer games. The characteristics of robot soccer environment were first analyzed. Then a multi-robot cooperation strategy was designed by referring to multi-agent theory. The strategy was built with the hierarchical structure of centralized hybrid system. The coordination layer constructed a real-time control system with the system architecture based on the symbol reasoning system, and established a set of rule library adapting to the robot soccer environment by analyzing and comparing some rules of condition analysis, formation control, role distribution and behavior selections with a lot of papers and experiments. During the running, real-time control system reasoned with rule libraries. The behavior layer built on behavior-based system architecture, divided behaviors into basic behaviors and complex behaviors, of which the latter comprised the former. The section of complex behaviors included some practical path planning methods. Moreover, in the thesis, the Q-learning algorithm in enforcement learning was introduced into the role distribution to get reasonable and valid distribution rules. Also, some problems of Q-learning application were found, which could be the revelation for the further study.A multi-robot cooperation program was designed according to the requirements of practical system. It controlled the multi-robot cooperation system in robot soccer environment with technologies of DLL, Object Oriented, complex data structures, data transfer, etc. In the experiments, the strategy developed in the thesis had advantages of good robust capabilities and rapid dealing speed, and could satisfy the requirements of robot soccer environment. The software run well in the Wulihe Cup International Robot Soccer Match...