自适应遗传算法和RBF网络在传球中的应用  

Application of Adaptive Genetic Algorithm and RBF Network in Passing Ball

作　　者：廖本先[1] 杨宜民[1] 张学习[1] 项凡[1] 

机构地区：[1]广东工业大学自动化学院,广东广州510090

出　　处：《计算机仿真》2010年第9期169-172,共4页Computer Simulation

摘　　要：机器人足球仿真比赛系统是研究人工智能的优秀平台,借助平台,将智能算法应用到仿真球队的相关策略设计中,通过球队之间的比赛以验证算法的可行性。传球是球员的基本动作之一,设计的好与否直接影响着球队的整体实力。由于仿真比赛环境是一个实时、动态、有干扰的环境,难以对传球动作建立精确的物理模型。为提高球队近似物理模型的传球成功率,提出了一种新的传球方法,即基于自适应遗传算法的RBF神经网络传球方法。用自适应遗传算法优化RBF神经网络的结构参数,通过优化,提高了网络的学习能力和全局搜索效率。仿真结果表明,经过自适应遗传算法优化的RBF神经网络的传球成功率得到了很大提高。The system of Robot soccer simulation game is the excellent platform. With the help of it,the intelligent algorithm is applied to the strategy design of simulation team and validate the feasibility of algorithm by competing with each other. In fact,kinds of intelligent algorithm are used widely to the design of team＇s fundamental action. Passing ball is one of the basic actions and impacts directly the team＇s entire efficiency. Since the environment of simulation game is a real-time,dynamic and interfered environment,it is difficult to establish an accurate physical model for the action of passing ball. In view of the ineffective of passing ball based on the past algorithm,the paper gives a new method of passing ball,which integrates the adaptive genetic algorithm and RBF Neural Network in passing ball,namely,using the adaptive genetic algorithm to optimize the structure parameter of RBF Neural Network. It enhances the network learning capability and improves global search efficiency of network. The experiment shows that the success rates of passing ball with RBF Neural Network optimized by adaptive genetic algorithm is improved greatly.

关 键 词：机器人足球 传球 自适应遗传算法 神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]