On the probabilistic feasibility of solutions in multi-agent optimization problems under uncertainty

TL;DR

This paper develops probabilistic guarantees for solutions in uncertain multi-agent optimization problems, focusing on constraint uncertainty and cost functions, with applications to electric vehicle charging control.

Contribution

It introduces distribution-free probabilistic robustness certificates for multi-agent optimization solutions, reducing sample complexity and applicable to various algorithms.

Findings

Provides agent-independent robustness bounds based on support rank.

Reduces sample complexity for probabilistic guarantees as the number of agents grows.

Demonstrates applicability through a case study on electric vehicle charging.

Abstract

We investigate the probabilistic feasibility of randomized solutions to two distinct classes of uncertain multi-agent optimization programs. We first assume that only the constraints of the program are affected by uncertainty, while the cost function is arbitrary. Leveraging recent a posteriori developments of the scenario approach, we provide probabilistic guarantees for all feasible solutions of the program under study. This result is particularly useful in cases where numerical difficulties related to the convergence of the solution-seeking algorithm hinder the exact quantification of the optimal solution. Furthermore, it can be applied to cases where the agents' incentives lead to a suboptimal solution, e.g., under a non-cooperative setting. We then focus on optimization programs where the cost function admits an aggregate representation and depends on uncertainty while constraints are deterministic. By exploiting the structure of the program under study and leveraging the so called support rank notion, we provide agent-independent robustness certificates for the optimal solution, i.e., the constructed bound on the probability of constraint violation does not depend on the number of agents, but only on the dimension of the agents' decision. This substantially reduces the number of samples required to achieve a certain level of probabilistic robustness as the number of agents increases. All robustness certificates provided in this paper are distribution-free and can be used alongside any optimization algorithm. Our theoretical results are accompanied by a numerical case study involving a charging control problem of a fleet of electric vehicles.