基于半马尔可夫对策的多机器人分层强化学习

在多智能体分层强化学习研究成果的基础上,考虑多机器人系统经常面临的通信困难问题(如水下环境),提出一种基于半马尔可夫对策的多机器人分层强化学习方法,通过引入对策论方法解决通信困难情况下多机器人学习问题。仿真实验结果表明了该方法的有效性。

Multi-robot hierarchical reinforcement learning based on semi-Markov games

Following previous work and considering the multi-robot systems with communication failure (e.g. in an underwater environment), a multi-robot hierarchical reinforcement learning approach based on semi-Markov games was proposed.  The game theory was employed in this approach. Simulation experimental results showed that the proposed approach was effective on multi-robot learning with communication failure.