摘要
针对于RoboCup比赛中出现的传球精度不够准确的问题,通过对决策树学习方法的探讨,该文提出了一种用于Robo-Cup仿真球队中Agent学习传球技能的一种决策树方法。将C5.0即ID3的改进算法应用到Agent传球能力的训练中,它使得Agent能够根据场上的具体情况,把球成功传给队友。Agent在得到球的控制权之后,首先确定传球成功率最大的球员,然后并不直接执行传球的动作,而是调整Agent自身的准备动作以达到传球的最佳状态,最后进行传球的行为。仿真结果表明,该方法有效地提高了Agent的传球能力。
The learning methods based on Decision Tree are discussed in order to solve the problem that passing ball in RoboCup games is not precise. A training algorithm for the ＂Passing Ball Ability＂ of the agents in the RoboCup Simulator League is presented in this paper. It makes agents determine that passing ball to which teammate is the best choice according to the current status in the field. When agent has controlled ball and determined to pass ball to which teammate, it does not pass ball right now, but it will adjust itself to reach the best condition in order to adapt the current situation. Finally it will do ＂passing ball＂ action. The results of the experiments have proved that this training algorithm really is good for passing ball.
出处
《计算机仿真》
CSCD
2006年第4期132-134,153,共4页
Computer Simulation
