Q学习算法在RoboCup带球中的应用

Application of Q-Learning Algorithm in Dribbling Ball Training of RoboCup

机器人世界杯足球锦标赛(RoboCup)是全球影响力最大的机器人足球比赛之一,而仿真组比赛是其重要的组成部分。鉴于带球技术在仿真组比赛中的重要性,我们将Q学习算法应用于带球技术训练中,使智能体本身具有学习和适应能力,能够自己从环境中获取知识。本文描述了应用Q学习算法在特定场景中进行1vs.1带球技术训练的方法和实验过程,并将训练方法应用于实际球队的训练之中进行了验证。

RoboCup is one of the most influential Robot soccer games,and the simulation game is an important part of it. Whereas the skill of dribbling ball has a very important place in RoboCup game, by trying to introduce Q-Learning method into the training of dribbling ball, we make the agent owning the ability to learn and adaptive to the environment, so that it can gain knowledge from the environment. This paper illustrates the Q-Learning algorithm in the special area 1 vs. 1 dribbling strategy train learning method and training process, the training method is applied to everyday training and its effect was validated.