Lien Zoom.

Potential games are a powerful tool in multi-agent decision-making: their special structure guarantees that natural learning dynamics converge to a Nash equilibrium. However, classical results on potential games are largely developed for finite action spaces, while many control and engineering applications, from distributed energy systems to multi-agent coordination, naturally arise in continuous state and action spaces. This talk presents recent results that characterize when linear-quadratic (LQ) games are potential. A first result establishes a structural limitation: in the two-player full-state-feedback setting, the set of LQ potential games is shown to be limited, essentially reducing to a small neighborhood of identical-interest games. Motivated by this, attention is restricted to LQ games with decoupled dynamics and decoupled linear state feedback. For this subclass, the set of potential games is shown to be richer; a characterization of the potential condition is established, the corresponding potential function is derived, and the existence of a Nash equilibrium is proved. The talk closes by discussing open challenges in characterizing Nash equilibria for this class, and connecting the results to a broader research agenda on learning in multi-agent dynamic systems.

Biography: Sara Hosseinirad is a recent PhD graduate from the University of British Columbia, Department of Electrical and Computer Engineering, where she was a Vanier Canada Graduate Scholar. Her research interests lie at the intersection of reinforcement learning, optimal control, and game theory, with a focus on principled methods for decision-making in multi-agent and partially observable systems. Her doctoral work includes theoretical contributions to multi-agent learning in dynamic games, as well as hierarchical reinforcement learning and model-predictive control architectures for safety-critical multiple-variable biomedical applications. She holds an MASc in Mechatronics from UBC and a BSc dual degree in Aerospace Engineering and Physics from Sharif University of Technology.