A Design & Implementation Of Strategy Training Software For Robot Soccer

Robot soccer is the product of the combination of soccer, robot and some other high technologies; one of its prospects is that a team of robots can win against the human world soccer champion team. The robot soccer system mainly includes four subordinate modules, which are robot dolly, visual, decision-making and communication module.Decision-making module works as the brain of the robot foot team, it plays a role of critical importance for the behavior of the whole team. There are plenty of methods to identify the strategy of a certain team, such as expert system, case learning, reinforcement learning and evolutionary algorithm.Genetic algorithm is an adaptive probabilistic searching algorithm for overall optimization which simulates the natural procedure of heredity and evolution and a robust searching algorithm which can be used to solve the optimization problem of complicated system. It is especially suitable for combination optimization purpose; besides, genetic algorithm can find the best answer automatically without manual intervention.The characteristics of genetic algorithm make it possible to be applied in the strategy training of a robot soccer team. This paper will present a method of strategy training based on the basic genetic algorithm and also describe the details of parameter choice, sufficiency evaluation, strategy division and chromosome coding methods with the mainly flow as well.Based on the method presented, the paper will give a description of the design and implementation of the strategy training program based on the basic genetic algorithm, including expectation, choice of environments and tools, user interface, design and implementation of software architecture and property browse characteristic.