Decentralized multi-agent reinforcement learning with networked agents: recent advances

Multi-agent reinforcement learning (MARL) has long been a significant research topic in both machine and control systems. Recent development of (single-agent) deep created resurgence interest developing new MARL algorithms, especially those founded on theoretical analysis. In this paper, we review recent advances sub-area topic: decentralized with networked agents. scenario, multiple agents perform sequential decision-making common environment, without the coordination any central controller, while being allowed to exchange information their neighbors over communication network. Such setting finds broad applications operation robots, unmanned vehicles, mobile sensor networks, smart grid. This covers several our endeavors direction, as well progress made by other researchers along line. We hope that promotes additional efforts exciting yet challenging area.