计算机工程与应用 ›› 2011, Vol. 47 ›› Issue (18): 53-55.

• 研究、探讨 • 上一篇    下一篇

基于Q学习Robocup前锋的射门训练

申 迅,刘国栋   

  1. 江南大学 通信与控制工程学院,江苏 无锡 214000
  • 收稿日期:1900-01-01 修回日期:1900-01-01 出版日期:2011-06-21 发布日期:2011-06-21

Robocup vanguard’s goal-scoring ability based on Q-learning

SHEN Xun,LIU Guodong   

  1. School of Communication and Control Engineering,Jiangnan University,Wuxi,Jiangsu 214000 China
  • Received:1900-01-01 Revised:1900-01-01 Online:2011-06-21 Published:2011-06-21

摘要: 3D仿真机器人是当今人工智能领域里一个极富挑战性的高技术密集型项目。在Robocup 3D比赛中,由于现今球队的人数较少,一个球队的进攻能力往往取决于前锋的个人能力,因此增强前锋的射门能力就显得十分重要。Q学习是一种重要的强化学习方法,将Q学习用到Robocup仿真环境中,使智能体通过在线学习获得射门技巧,并且通过实际比赛证明了算法的有效性。

关键词: Robocup, Q学习, 智能体, 射门

Abstract: 3D simulation of robot soccer is one of the modern challenging high-tech-intensive projects in artificial intelligence field.Due to the small number of players in one team in the competition,the offensive capability of a team often depends on the vanguard ’s personal abilities,and it is extremely necessary to enhance ability of vanguard’s shot ability.Q-learning is an important method of reinforcement learning.This paper uses Q-learning in simulation environment,so that agent can learn shot skills through the online learning,and it has proved the effectiveness of the algorithm through the actual game.

Key words: Robocup, Q-learning, Agent, shot