Distributed Learning, Extremum Seeking, and Model-Free Optimization for the Resilient Coordination of Multi-Agent Adversarial Groups

Abstract : This proposal focuses on the analysis and design of coordination algorithms for multiple agents deployed in adversarial environments. The multi-agent systems can represent miscellaneous autonomous and semi-autonomous vehicles that are remotely controlled by operators. These groups can be subject to attacks from other external agents leading to complex networked adversarial settings. The proposal presents work in two main areas: 1) the use of a class of receding-horizon type of algorithms to overcome the effect of a type of uncoordinated attackers on a multi-vehicle-operator group, and 2) the use of extremum seeking (ES) techniques to learn Nash equilibria in finitely- and infinitely-many player noncooperative games and to solve high-dimensional optimization problems. Extensions and applications of these techniques were developed during the realization of the project.