Study On The Mothed Of Multi-Robot Cooperation Based On The Multi-Agent System

Multi-robot system has been paid more and more attention and research because of its incomparable advantages over single robot. Studying based on the multi-agent system is a most adopted idea by robotic researchers. After detailed understanding and long-term study of multi-agent system, this thesis discussed and realized a multi-agent robot system, KD-MARCS.The system is composed of four pioneer2 robots and a few personal computers, the whole system can be seemed as a Hierarchical Agent Network. Then we defined a composite communication model and a message-based protocol.We presented two collaborative localization(CoLocalization) methods to solve the robot localization problem, one is based on the sonar ring, the other is based on the camera and laser range finder. The former has easy realization but poor precision, whereas the latter has a little more complexity but even high precision.As to the task allocation problem, we choosed a simplified "English Auction", that is, the highest price, public bid and once negotiation model to deal with the problem, and we presented the auction protocol. To describe different tasks, we presented and realized an idea of "Hierarchical Task Structure".At the end of the thesis, we brought out the system control software, and described a distance adjustment experiment, a delivery experiment and a push-box experiment. The results showed that the CoLocalization based on camera and laser range finder can provide high-precision localization, and the task allocation method based on the auction model could effectively solve the resource allocation problem and schedule the robots reasonably.