seminars

Last Tuesday during our group seminar, we hosted Ye Li — an assistant professor from Beijing University of Technology currently visiting Jagiellonian University for three months through a collaboration supported by the European Research Council and National Natural Science Foundation of China. Ye specializes in Macroscopic Fundamental Diagram (MFD) research, and together we are investigating new approaches to routing autonomous vehicle fleets within MFD frameworks. The methodology looks very promising, and we hope to share results soon. 👌

After a longer break, we have returned from the summer holidays to our tradition of seminar meetings. Today at our Kraków lab, we had the pleasure of hosting Julian Sienkiewicz from Warsaw University of Technology, who delivered a seminar: Agent-based modeling of Multilevel Information Overload: assumptions and expectations. Many thanks to our guest for an incredibly interesting presentation, and to all team members for their presence and engaging discussion.

Fryderyk Falniowski presented his work titled Heterogeneity, Reinforcement Learning and Chaos in Population Games. Abstract: Inspired by the challenges at the intersection of Evolutionary Game Theory and Machine Learning, we investigate a class of discrete-time multi-agent reinforcement learning (MARL) dynamics in population/non-atomic congestion games, where agents have diverse beliefs and learn at different rates. These congestion games, a well-studied class of potential games, are characterized by individual agents having negligible effects on system performance, strongly aligned incentives, and well-understood advantageous properties of Nash equilibria. Despite the presence of static Nash equilibria, we demonstrate that MARL dynamics with heterogeneous learning rates can deviate from these equilibria, exhibiting instability and even chaotic behavior and resulting in increased social costs. Remarkably, even within these chaotic regimes, we show that the time-averaged macroscopic behavior converges to exact Nash equilibria, thus linking the microscopic dynamic complexity with traditional equilibrium concepts. By employing dynamical systems techniques, we analyze the interaction between individual-level adaptation and population-level outcomes, paving the way for studying heterogeneous learning dynamics in discrete time across more complex game scenarios.

Anastasia presented and overviewed the broad field of Communication in MARL.

Hosting Paweł Gora who presented his works, including traffic lights optimization in large scale simulator that he has implemented.

Meeting with Zafeira Gkioulou and discussion on the future of LajkBus.