A Roadmap Towards Improving Multi-Agent Reinforcement Learning With Causal Discovery And Inference
以因果发现与推断改进多智能体强化学习的研究路线
Causal reasoning is increasingly used in Reinforcement Learning (RL) to improve the learning process in several dimensions: efficacy of learned policies, efficiency of convergence, generalisation capabilities, safety and interpretability of behaviour. However, applications of causal reasoning to Multi-Agent RL (MARL) are still mostly unexplored. In this paper, we take the first step in investigating the opportunities and challenges of applying causal reasoning in MARL. We…