Algorithmic Collusion among EV Charging Stations with Independent Reinforcement Learning Agents
Publication Date
10-21-2025
Document Type
Conference Proceeding
Publication Title
2025 IEEE International Conference on Communications Control and Computing Technologies for Smart Grids Smartgridcomm 2025 Proceedings
DOI
10.1109/SmartGridComm65349.2025.11204603
Abstract
Recent advancements in reinforcement learning (RL) have significantly enhanced the autonomous decision-making capabilities of energy systems. In the realm of electric vehicle (EV) charging stations, operators aim to autonomously learn optimal pricing strategies based on local observations and user behavior. Traditional multi-agent reinforcement learning (MARL) approaches often assume cooperative settings with reward sharing or communication, real-world EV charging infrastructures are typically decentralized and competitive. This paper investigates the potential for tacit collusion in such realistic, non-cooperative environments. By leveraging decentralized MARL, we show that independently trained agents can still exhibit emergent collusive behaviors. Our findings highlight a critical need for systematic evaluation and detection of algorithmic collusion in autonomous pricing systems.
Funding Number
70NANB24H301
Funding Sponsor
Concordia University
Department
Mechanical Engineering
Recommended Citation
Yulin Zeng, Yiheng Zhao, Yuanliang Li, Yefeng Yuan, Jie Gao, Hepeng Li, Xiao'Ou Yang, Hohyun Lee, Yuhong Liu, and Jun Yan. "Algorithmic Collusion among EV Charging Stations with Independent Reinforcement Learning Agents" 2025 IEEE International Conference on Communications Control and Computing Technologies for Smart Grids Smartgridcomm 2025 Proceedings (2025). https://doi.org/10.1109/SmartGridComm65349.2025.11204603
