Operations Research Seminar: Maryam Kamgarpour

Multi-agent decision-making in uncertain environments

Maryam Kamgarpour, ETH Zurich

Multi-agent decision making arises in several applications ranging from electricity markets to telecommunication and transportation networks. In these settings, each agent aims to optimise its objective function while ensuring local and global constraints. Hence, the agents need to play a feasible Nash equilibrium to ensure a stable solution outcome. A challenge in designing an algorithm to find the Nash equilibrium is that each agent’s cost function depends in a non-trivial way on other agents’ strategies. The agents can only measure their costs for a given action rather than knowing its functional form. I will describe our decentralized approach to learn Nash equilibria in a convex game.
Furthermore, I will illustrate the applicability of the approach to a game arising in electricity markets. The talk is based on joint work with Tatiana Tatarenko and parts of the results are presented in the following paper.

T. Tatarenko and M. Kamgarpour, Learning Generalized Nash Equilibria in a Class of Convex Games, IEEE Transactions on Automatic Control, 2019, to appear, available at https://arxiv.org/abs/1703.04113