Three Perspectives on Multi-Agent Reinforcement Learning

Abstract

This chapter concludes three perspectives on multi-agent reinforcement learning (MARL): (1) cooperative MARL, which performs mutual interaction between cooperative agents; (2) equilibrium-based MARL, which focuses on equilibrium solutions among gaming agents; and (3) best-response MARL, which suggests a no-regret policy against other competitive agents. Then the authors present a general framework of MARL, which combines all the three perspectives in order to assist readers in understanding the intricate relationships between different perspectives. Furthermore, a negotiation-based MARL algorithm based on meta-equilibrium is presented, which can interact with cooperative agents, games with gaming agents, and provides the best response to other competitive agents.