论文部分内容阅读
为了提高多智能体系统中的典型的强化学习——Q学习的学习效率和收敛速度,充分利用环境信息和相关的专家经验,提出了具有先验知识的Q学习算法。利用模糊综合决策方法处理专家经验和环境信息得到Q学习的先验知识,对Q学习的初始状态进行优化。典型的足球机器人系统中的仿真实验结果表明:该算法使学习过程建立在较好的学习基础上,从而更快地趋近于最优状态,其学习效率和收敛速度明显优于普通的Q学习。
In order to improve the learning efficiency and convergence speed of typical intensive learning in multi-agent system and make full use of environmental information and relevant expert experience, a Q learning algorithm with prior knowledge is proposed. The fuzzy comprehensive decision-making method is used to process the expert’s experience and environmental information to obtain the prior knowledge of Q learning and to optimize the initial state of Q learning. The simulation results of a typical soccer robot system show that the algorithm makes the learning process based on better learning and approaches the optimal state faster, and its learning efficiency and convergence speed are obviously better than those of ordinary Q learning .