Based On The Mas Multi-robot Soccer Simulation Platform For The Realization Of The Collaborative Tactics

Multi-Robot Soccer, an interesting and complex new area of Distributed Artificial Intelligence, has attracted increasing attention in recent years. Multi-Robot Soccer is a typical Multi-Agent System, in which both the player and the ball can be viewed as agents. This paper focuses on the decision-making and reasoning of single robot, cooperating mechanism of team and local collaboration, in dynamic and unpredictable environment of game. In real simulation game, bottom level survival behavior, mid-level strategy behavior and top-level game tactics and strategy, these three levels functions guarantee that the game can be performed continuously.On the basis of the Multi-Agent System theory, this paper divides single robot's behavior into survival behavior and strategy behavior. Limited actions set has been made for every robot based on these two behaviors. Whereafter, single robot's decision-making subsystem has been designed. Every robot selected and performed desired actions by the strategy forward reasoning, basing on the tasks assigned by the system. An behavior-selection mechanism, based on practicability and tendency was proposed by this paper. This mechanism has been proved to be a practical and effective decision-making method.Cooperating mechanism among soccer robots in the game includes individual competition, local strategy, global strategy and some other cooperation and negotiation. The dilemma of making decision when individual and team have different goals has also been studied. This paper researches teamwork cooperation tactics and strategies basing on the Multi-Agent negotiation. For teamwork and its formation, the method of object-oriented programming is used to analyze the roles of ballplayers of robot team, and the corresponding tactics and strategies are designed. Paper introduced a cooperating mechanism of supervising the group behaviors based on coach agent.On the simulation platform of Multi-Robot Soccer, two different tactics, named Attack-Defend Equation Strategy and Attack-Defend Entirely Strategy had been arranged into two team formations to counterwork each other. By finding out the reason of Attack-Defend Entirely Strategy unsuccessful score, paper introduces an improved Attack-Defend Entirely Strategy based on triangle-location. Simulation results show that this improved strategy improves greatly in solving the conflict among agents and results in the real-time and the flexible for robots. It can be a way of reference solution for further study.