弱通信条件下基于Q图迁移的多智能体分层强化学习
首发时间:2013-02-02
摘要:分层强化学习是为解决强化学习的维数灾问题而提出的,在多智能领域已有推广,但是在弱通信受限条件下不能有效地解决实际问题。本文在现有多智能体分层强化学习研究成果的基础上,引入半马尔可夫对策模型及求解算法,并构造了Q图迁移算法,从而提出一种弱通信条件下基于Q图迁移的的多智能体分层强化学习方法,仿真实验结果表明了该方法的有效性和优越性。
For information in English, please click here
Multi-Agent Hierarchical Reinforcement Learning Based on Q-Graph Transferring with Weak Communication
Abstract:The approaches of hierarchical reinforcement learning (HRL) are proposed for combating the curse of dimensionality bedeviling reinforcement learning. Some HRL approaches have been applied in multi-agent systems. But these approaches cannot efficently deal with under the conditions of weak communication. Based on the previous works, a multi-agent HRL approach based on Q-graph transferring under the conditions of weak communication is proposed by employing the model and solving algorithm of semi-Markov game, and Q-graph transfer algorithm. The simulation expremental results show that the proposed approach is effective and advantaged.
Keywords: weak communication, multi-agent system, hierarchical reinforcement learning, transfer learning
论文图表:
引用
No.****
同行评议
共计0人参与
勘误表
弱通信条件下基于Q图迁移的多智能体分层强化学习
评论
全部评论0/1000