【摘要】 强化学习和规划技术在目标上有着很高的相似性,而在技术上又具有互补性,因此,基于强化学习的Agent规划规则抽取问题长期以来一直是研究的热点。针对基于强化学习的多Agent系统在规划规则抽取方面存在的问题,提出了一种从多Agent Q学习中抽取满足规划条件的规划规则的RL-MAPRE算法,并给出了理论分析。更多还原

【Abstract】 As two important aspects of artificial intelligence research,the reinforcement learning and planning method are very similar in retrieving goals,while they are also inter-complementary in the technique.Therefore,taking planning rules from reinforcement learning based system has been a hot spot for a long time in the field.This paper proposes a RL-MAPRE algorithm,which extracts the planning required rules in the multi-agent Q-learning based on Nash equilibrium.更多还原