机器人足球智能体行为选择策略仿真研究  

Selecting Strategy for Agent Behavior Based on Fuzzy Algorithm and Q-learning

作　　者：吕家杰[1] 王改云[1] 

机构地区：[1]桂林电子科技大学,广西桂林541004

出　　处：《计算机仿真》2012年第9期174-177,共4页Computer Simulation

摘　　要：研究足球机器人供过传球的成功率,优化控制策略,在机器人足球比赛中,针对现有行为选择策略因不能综合全面地考虑场上复杂的影响因素,造成传球、带球和射门的失误率较高的问题。为提高球员带球的准确率,提出了一种模糊Q学习的行为选择决策机制。利用模糊算法全面地处理和评估场上的干扰和影响因素,做出最佳行为决策。并通过Q学习算法来修正模糊规则。采用上述算法的行为选择策略,增强了策略对动态环境的适应能力。在机器人足球世界杯2D仿真平台上对策略进行了仿真验证,仿真结果表明改进算法能够很好的改善智能体的射门、传球和带球的成功率。This paper put forward a decision-making mechanism for behavior selection based on fuzzy Q learning. The fuzzy algorithm was used to evaluate and process the factors which impact the player＇ s behavior on the field. Simultaneously, the Q-learning algorithm was used to modify the fuzzy rules. The resuhs of the experimental with the simulation platform show that this algorithm is efficient and can improve the success rates of robot player in shotting, passing and dribbling.

关 键 词：模糊算法 加强学习 行为选择 决策 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]