竞争与合作视角下的多Agent强化学习研究进展OA北大核心CSTPCD

With the rapid development of deep learning and reinforcement learning,multi-agent reinforcement learning(MARL)has become a common approach to solve the large scale complex sequential decision-making problem.In order to promote the development of this field,this paper collects and reviews recent research results from the perspective of competition and cooperation.This paper introduced deep reinforcement learning and introduced the basic theoretical framework of MARL-Markov game and extensive game,and especially emphasized the reinforcement learning algorithms developed recently in three scenarios of competition,cooperation and mixture.This paper discussed the core challenge of MARL that was non-stationary of the environment,and an example was given to summarize and prospect its solutions.