Advances and Applications in Multi-Agent Reinforcement Learning

In today's life, reinforcement learning is applied to many fields and has shown good results, such as driverless cars and Warehouse sorting goods robots. From the beginning of simple single-agent intelligent control, researchers gradually turn their attention to multi-agent cooperative games to solve production problems. Compared with the single-agent training algorithm, multi-agent reinforcement learning is more complex and difficult to converge. Therefore, it is particularly important to study the progress and application of multi-agent reinforcement learning. This paper reviews three perspectives, which are the introduction of reinforcement learning algorithms, the application scenarios of multi-agent reinforcement learning (games, path planning, and robot cooperation), the problems faced, and future prospects. It is hoped that this review can provide researchers with more comprehensive information for subsequent in-depth research. At the same time, this paper puts forward more innovative ideas, hoping that researchers can expand their ideas and promote the development of multi-agent reinforcement learning.