Abstract

The rapid progress of intelligent systems has heightened interest in reinforcement learning (RL), which enables autonomous agents to acquire optimal behaviors through interaction with their environments. As tasks and environments grow more complex, the demand for agents that can act both independently and collaboratively has become increasingly evident. This has given rise to multi-agent reinforcement learning (MARL), a field dedicated to developing frameworks where multiple agents can cooperate, compete, or coexist to achieve individual or collective goals. MARL is particularly relevant in real-world domains such as robotics, autonomous driving, and finance, where agents must adapt their decisions based not only on their own actions but also on those of others. This talk covers the challenges and opportunities posed by MARL in practical applications, where agents face partial observability, non-stationarity, and the necessity of coordinated decision-making. While traditional single-agent reinforcement learning (SARL) approaches perform well in isolated environments, they are insufficient in multi-agent settings due to the added complexity of inter-agent dynamics. These challenges call for advanced algorithms that support decentralized decision-making, foster cooperation, and ensure robustness. In this talk, we discuss novel methods designed to improve the efficiency and effectiveness of MARL systems in both discrete and continuous action spaces.

About the Speaker

Reza Azadeh is an Associate Professor with the Miner School of Computer and Information Sciences at the University of Massachusetts Lowell (UML), where he directs the Persistent Autonomy and Robot Learning (PeARL) lab. His research interests encompass robot learning and autonomy. He is the recipient of an NSF CAREER Award and a senior member of IEEE. Before joining UML, he was a Postdoctoral Fellow with the School of Interactive Computing at Georgia Institute of Technology. He holds a Ph.D. in Robotics, Cognition, and Interactive Technologies from the University of Genoa, in collaboration with the Italian Institute of Technology.