Reinforcement learning (RL), as one of computational-intelligence approaches, has been proved to be a powerful tool to solve complicated decision-making problems, e.g., the success of AlphaGo and AlphaStar. In real world, many systems are composed of multiple agents that interact under the same environment with each other. They may cooperate to achieve better results compared with the case that single agent is involved, or may compete with each other for finite resources. Motivated by that, multi-agent reinforcement learning (MARL) has attracted more and more attention from both academy and industry. In this special session, we would like to invite worldwide researchers to share and present their latest work of MARL on theory, methodology, and applications, and discuss future directions for MARL.
Multi-agent systems (MAS) can be found in many real-world areas, including financial market, military systems, multi-robot systems, cyber-physical systems, transportation systems, sensor networks, smart grids, IT systems, factory automation, biological systems and so on. In such systems, interactions among agents and environment can be increasingly complicated due to the individual behaviors of each agent. In order to achieve satisfying cooperative or competitive performance, MARL has to fully overcome the challenges including credit assignment, non-stationarity, combinatorial explosion, equilibrium convergence, and so on. It requires latest advances in both theory and methodology. In addition, applications of MARL in practical systems and real-world environment are critical for the development of the area. It has to fully consider the control/communication limit of internal systems and the noise/disturbance from external environment.
IEEE WCCI 2022 is the world’s largest technical event on computational intelligence (CI), and from either academic point or practical significance, MARL is an important topic for the development of CI. We hope the organization of this special session in IEEE WCCI 2022 would provide an opportunity for researchers from both academy and industry over the world to share and present their latest research in the theory, methodology, and applications of MARL. It will predictably prompt fresh thinking on the future directions and encourage more cooperation among researchers.
Submissions on all aspects of MARL, including, but not limited to, the following topics are welcomed:
Multi-agent reinforcement learning algorithms
Cooperation/competition/mixed cooperation-competition in MAS
Centralized/distributed training and centralized/decentralized execution
Emergent patterns of agent behaviors
Methodology of communication/coordination/scaling/credit assignment in multi-agent systems
Non-stationarity/partial observability analysis
Theory of multi-agent optimality/Nash equilibrium/convergence stability/etc.
Multi-objective optimization of MAS
Applications of MARL in practical and large-scale systems
Paper submission: January 31st, 2022 (11:59 PM AoE)
Notification of acceptance: April 26, 2022
Final paper submission: May 23, 2022
Papers submitted to this Special Session are reviewed according to the same rules as the submissions to the regular sessions of WCCI 2022. Authors who submit papers to this session are invited to mention it in the form during the submission. Submissions to regular and special sessions follow identical format, instructions, deadlines and procedures of the other papers.
Please, for further information and news refer to the WCCI website: https://wcci2022.org/
Yuanheng Zhu
Chinese Academy of Sciences, Institute of Automation
Zongqing Lu
Peking University
Jianye Hao
Tianjin University