Distributed extremum seeking for real-time resource allocation

In this paper we present a distributed extremum seeking control, designed to perform real-time resource allocation in multi-agent systems (MAS). The algorithm allows the optimal allocation of a given resource, in a distributed architecture, where each agent shares information only with its neighboring agents, and needs to only measure its own individual payoff function. We combine ideas from recent evolutionary game theory dynamics and the classic extremum seeking control to ensure convergence to an optimal allocation. The application of the algorithm is illustrated via simulation.