Algorithm design and approximation analysis on distributed robust game

TL;DR

This paper introduces a distributed algorithm to find approximate generalized Nash equilibria in robust games with uncertain constraints, addressing the challenge of high-dimensional nonlinear uncertainties through polyhedral approximation and robust optimization.

Contribution

It proposes a novel distributed method for robust game equilibrium seeking under uncertainty, with convergence guarantees and approximation error analysis.

Findings

Algorithm converges to an $ ext{epsilon}$-Nash equilibrium

Approximation accuracy depends on problem parameters

Provides bounds on the equilibrium approximation error

Abstract

We design a distributed algorithm to seek generalized Nash equilibria of a robust game with uncertain coupled constraints. Due to the uncertainty of parameters in set constraints, we aim to find a generalized Nash equilibrium in the worst case. However, it is challenging to obtain the exact equilibria directly because the parameters are from general convex sets, which may not have analytic expressions or are endowed with high-dimensional nonlinearities. To solve this problem, we first approximate parameter sets with inscribed polyhedrons, and transform the approximate problem in the worst case into an extended certain game with resource allocation constraints by robust optimization. Then we propose a distributed algorithm for this certain game and prove that an equilibrium obtained from the algorithm induces an $\epsilon$-generalized Nash equilibrium of the original game, followed by convergence analysis. Moreover, resorting to the metric spaces and the analysis on nonlinear perturbed systems, we estimate the approximation accuracy related to $\epsilon$ and point out the factors influencing the accuracy of $\epsilon$.