Multi-agent reinforcement learning for character control

Abstract Simultaneous control of multiple characters has been a research topic that extensively pursued for applications in computer games and animations, such as crowd simulation, controlling two carrying objects or fighting with one another team playing collective sports. With the advance deep learning reinforcement learning, there is growing interest applying multi-agent intelligently to produce realistic movements. In this paper we will survey state-of-the-art MARL techniques are applicable character control. We then papers make use multi-character discuss about possible future directions research.