A survey on multi-agent deep reinforcement learning: from the perspective of challenges and applications
- PDF / 1,108,811 Bytes
- 24 Pages / 439.37 x 666.142 pts Page_size
- 60 Downloads / 225 Views
A survey on multi‑agent deep reinforcement learning: from the perspective of challenges and applications Wei Du1 · Shifei Ding1,2
© Springer Nature B.V. 2020
Abstract Deep reinforcement learning has proved to be a fruitful method in various tasks in the field of artificial intelligence during the last several years. Recent works have focused on deep reinforcement learning beyond single-agent scenarios, with more consideration of multiagent settings. The main goal of this paper is to provide a detailed and systematic overview of multi-agent deep reinforcement learning methods in views of challenges and applications. Specifically, the preliminary knowledge is introduced first for a better understanding of this field. Then, a taxonomy of challenges is proposed and the corresponding structures and representative methods are introduced. Finally, some applications and interesting future opportunities for multi-agent deep reinforcement learning are given. Keywords Deep reinforcement learning · Multi-agent · Game theory · Centralized training and decentralized execution · Communication learning · Agent modeling
1 Introduction Reinforcement Learning (RL) is often considered to be a general formalization of decision-making tasks and a subfield of machine learning. In RL, agents learn not from sample data, as in supervised and unsupervised learning, but from experiences that interact with the environment. With the success of deep neural networks (DNN), reinforcement learning algorithms combine with it and form deep reinforcement learning (DRL) methods to solve complex problems in the real world. The pioneering model is Deep Q-Network, which was able to play Atari console games without adjusting network architecture or hyperparameters. Deep reinforcement learning methods have been extensively researched and significantly improved since then. Most successful DRL methods have been in the single-agent domains so far, and extending DRL to multi-agent settings is indispensable. However, deep reinforcement learning * Shifei Ding [email protected] 1
School of Computer Science and Technology, China University of Mining and Technology, Xuzhou 221116, China
2
Mine Digitization Engineering Research Center of Ministry of Education of the People’s Republic of China, Xuzhou 221116, China
13
Vol.:(0123456789)
W. Du, S. Ding
for multi-agent settings is fundamentally more difficult than the single-agent scenario due to the presence of multi-agent pathologies such as the curse of dimensionality and multiagent credit assignment. Despite this complexity, there has been a lot of work in the fields of general control, robot system (Gu et al. 2017; Kurek and Jakowski 2016), man–machine game (Fu et al. 2019; Lanctot et al. 2017; Leibo et al. 2017), autonomous driving (ShalevShwartz et al. 2016), Internet advertising (Jin et al. 2018), and resource utilization (Xi et al. 2018; Perolat et al. 2017). This paper systematically summarizes several research directions in the field of multiagent deep reinforcement learning (MDRL), including
Data Loading...
