Signal Learning with Messages by Reinforcement Learning in Multi-agent Pursuit Problem

Abstract

Listen

Translate article

Communication is a key for facilitating multi-agent coordination on cooperative problems. Reinforcement learning is one of the learning methods for such cooperative behavior of agents. Kasai et al. proposed Signal Learning (SL) and Signal Learning with Messages (SLM) by which agents learn policies of communication and action concurrently in multi-agent reinforcement learning framework. In this study, we experimented that the performance of the SLM is better than SL to pursuit problem where agents can observe only partial information and can move four directions. As a result, it has been shown that learning performance in SLM with longer messages is better than SL.