Prof. Shinkyu Park, Professor at the KAUST, visited CDSL and gave a seminar on “Learning Models in Large-Population Games.”
Unlike traditional game-theoretic approaches that assume static payoffs and full information, Prof. Park’s research considers dynamic payoff mechanisms and learning from instantaneous feedback. Using passivity-based analysis methods from control theory, he showed how learning dynamics can be designed to guarantee convergence to the Nash equilibrium, even when time delays are present. A higher-order learning model was also introduced to strengthen these convergence guarantees. The talk then turned to applications. In multi-robot task allocation, decentralized learning enabled groups of robots to coordinate effectively in dynamic environments. In another example, epidemic modeling was addressed through the design of dynamic payoff mechanisms in SIRS models, helping mitigate endemic transmission under stochastic influences.
The seminar was highly informative and engaging, providing CDSL students and researchers with valuable insights into the evolving interests in game theory and multi agent systems.
We thank Prof. Park for sharing his deep insights and for stimulating discussions on connection between game theory and control theory and look forward to potential collaborations in this critical area of research.