This talk examines the asymptotic behaviour of the price of anarchy as a function of the total traffic inflow in non-atomic congestion games with multiple origin-destination pairs. It first shows that the price of anarchy may remain bounded away from 1, even in simple three-link parallel networks with convex cost functions. On the other hand, empirical studies show that the price of anarchy is close to 1 in highly congested real-world networks, thus begging the question: Under what assumptions can this behaviour be justified analytically? To that end, it proves a general result showing that for a large class of cost functions (defined in terms of regular variation and including all polynomials), the price of anarchy converges to 1 in the high congestion limit. In particular, specialising to networks with polynomial costs, it shows that this convergence follows a power law whose degree can be computed explicitly.
Marco Scarsini is a Professor in the Department of Economics and Finance at LUISS, Rome, Italy. He obtained his Laurea in Economics and Social Sciences at Università Bocconi and his HDR in Applied Mathematics and Applications of Mathematics at Université Paris Dauphine. He had previous appointments in various universities, including SUTD. He has written over one hundred papers and serves on the editorial board of several scientific journals. His main research areas are applied probability and game theory, with a particular focus on congestion games and social learning.
For more information about the ESD Seminars Series, please contact Ying Xu at email@example.com.