Multi-robot Cooperation For Robot Soccer System Modeling With Petri Nets

Multi-robot system is a new aspect of robotics development. In the research of multi-robot system, multiple mobile robots coordination is a well-attended and essential topic. One of the typical multi-robot cooperation systems is Robot soccer system, and it is a standard platform for researching in multi-robot cooperation problems. This paper did some research on multi-robot cooperation, and chose a Vision-Based Remote Brainless Soccer Robot system as the experiment platform.Firstly, a three-layer structure of the collaborative, coordinate and behavior-control was proposed base on existing decision-making model and the research for the multi-robot control architecture under soccer environment, which involves some problems such as strategy and formation choice, role assignment and action selection. All kind of actions were designed and implemented in the action layer based on the analysis of the robot motion model, and this is the basis of the subsequent work.Secondly, we designed the coordination layer, which included the information pretreatment,the situation analyses,the strategy and formation chosen mechanism, the corresponding Petri net model was built based on it and its characters were analyzed by theory and simulator. The defend strategy of goalkeeper was designed independently to make this system follow a fixed goalkeeper scheme. The Petri net model of goalkeeper's strategy choice was built and analyzed theoretically, and emulational experimentations were done to verify the model's correctness and rationality.Finally, a role assignment method was presented in the coordination layer which was based on the role evaluating function of dynamic reference point, and the potential local conflict of the same kind of roles was solved by using the method of optimize estimate. The emulational experimentations approved that the method was simple and fast, and the task assignment was clearer. Taking the attack order for example, a Petri net model of task assignment was built based on this, and completed the analysis of this model. At last, a program framwork was given, and the effect of experimentations were analysed.