On logit dynamics in population games

We study the logit evolutionary dynamics in population games. For general population games, we prove that, on the one hand strict Nash equilibria are locally asymptotically stable under the logit dynamics in the low noise regime, on the other hand a globally exponentially stable fixed point exists in the high noise regime. This suggests the emergence of bifurcations in population games admitting multiple strict Nash equilibria, as verified numerically in previous publications. We then prove sufficient conditions on the game structure for global asymptotic stability of the logit dynamics in every noise regime. Our results find application in particular to heterogeneous routing games, a class of non-potential population games modelling strategic decision-making of users having heterogeneous preferences in transportation networks. In this setting, preference heterogeneities are due, e.g., to access to different sources of information or to different trade-offs between time and money. We show that if the transportation network has parallel routes, then the unique equilibrium of the game is globally asymptotically stable.