Distributed Online Learning for Leaderless Multicluster Games in Dynamic Environments
Rui Yu, Min Meng, Li Li
Abstract
In this paper, multi-cluster games with coupled constraints in dynamic environments are considered. Agents are divided into disparate clusters and each cluster can be regarded as a virtual noncooperative player competing against each other via a connected graph. The agents in a cluster are the actual players that aim to cooperatively minimize the cluster's time-varying payoff function, subject to local feasible set constraints and time-varying coupled inequality constraints. In this setting, a novel distributed online learning algorithm based on mirror descent and primal-dual approaches is devised to seek the generalized Nash equilibrium of the considered games. In addition, it can be proved that sublinearly bounded dynamic regrets and constraint violation can be guaranteed by felicitously choosing decreasing stepsizes in the proposed algorithm if the path length accumulation and gradient length accumulation increase sublinearly. Finally, a numerical simulation is presented to illustrate the theoretical results in this paper.