Sample and Communication Efficient Fully Decentralized MARL Policy Evaluation via a New Approach: Local TD update

arXiv:2403.15935v1 Announce Type: new
Abstract: In actor-critic framework for fully decentralized multi-agent reinforcement learning (MARL), one of the key components is the MARL policy evaluation (PE) problem, where a set of $N$ agents work cooperatively to evaluate the value function of the global states for a given policy through communicating with their neighbors. In MARL-PE, a critical challenge is how to lower the sample and communication complexities, which are defined as the number of training samples and communication rounds needed …

abstract actor actor-critic agent agents arxiv communication components cs.lg decentralized evaluation framework function global key multi-agent policy reinforcement reinforcement learning sample set the key type update value via work