Research On The Key Technology Of Decision-Making System For RoboCup Middle-Size Robot

In order to satisfy the human needs better, develop intelligent robot with human-like thinking mode has been urgently required in recent years. Nowadays RoboCup Middle-size soccer robot has become a good research platform for the intelligent robot, owing to its complete independent perception ability, independent decision-making ability and independent control ability.RoboCup Middle-size soccer robot has four systems:they are visual system, communication system, decision-making system and motion control system. Decision-making system is the core system which can give the robot artificial intelligence. This paper will make it as the study object, and do some research into its key technology.Firstly, this paper proposed a decision-making model to satisfy the requirements of the RoboCup Middle-size League competition, considering the characteristics of this competition. The model divided the soccer robot decision-making system into two layers. One layer is used for group decision-making, which completed by the outside trainer-computer. The other layer is the autonomy decision-making layer, and it's completed by the soccer robot itself. The main task of group decision-making is to assess the situation, furthermore assign the formation and role. And the autonomous decision-making layer basically completes those tasks such as behavior choice, path planning and the trajectory tracking. This paper wills build the decision-making system for the RoboCup Middle-size soccer robot on the basis of this model.Secondly, this paper chose eight factors which can objectively reflect the game situation, in addition using fuzzy comprehensive evaluation method to evaluate game situation, furthermore chose the formation based on the evaluation results and the number of effective robot in the competition field. What's more, using fuzzy comprehensive evaluation method again to evaluate the offensive ability or defensive ability of each robot, and assign a role for each robot on the basis of the evaluation results. Eventually this kind of group decision-making method was simulated in the matlab environment, and the result suggests this method is effective and feasible.Thirdly, according to the movement features of the RoboCup Middle-size soccer robot and the characteristics of RoboCup Middle-size League competition, this paper designed some basic and complex actions for the soccer robot. Then the behavior selection strategies were planed base on decision tree method, to ensure each robot can choose suit behavior when it acquired the current situation information and its own role.Finally, the paper proposed an optimal path planning algorithm base on Bezier curve mode. This algorithm choose the Bezier curve to describe robot path; and aim to gain the shortest path for the robot moving; besides it considers some factors such as the extreme speed, the extreme acceleration and obstacles, which could affect the robot motion. At last, the optimal path planning algorithm was simulated in the matlab simulation environment, and the result suggests this algorithm achieve an ideal effect.