摘要
提出了一种足球机器人基于Q学习与案例学习(CBL)相结合的自主学习机制。足球机器人通过给定的当前位置和奖赏信号,自己学习来实现特定的动作,为了降低学习时的计算复杂度,状态空间通过分段映射为不同的类别,根据其普遍性以及有限的所必须的计算机内存,采用神经网络的学习来执行其动作。
In this paper, a mechanism of behavior learning for soccer robot action selection based on Q learning and case based learning is proposed. The robot learns to activate a particular movement through their current given situation and rewarded signal. In order to decrease the number of stateaction pairs, the state space is segmented into different categories. Neural network is adopted to implementations of learning for their generalization properties and limited computer memory requirements.
出处
《西华大学学报（自然科学版）》
CAS
2005年第4期58-60,共3页
Journal of Xihua University:Natural Science Edition
