Learning generalized Nash equilibria in multi-agent dynamical systems via extremum seeking control

In this paper, we consider the problem of learning a generalized Nash equilibrium (GNE) in strongly monotone games. First, propose semi-decentralized and distributed continuous-time solution algorithms that use regular projections first-order information to compute GNE with without central coordinator. As second main contribution, design data-driven variant former algorithm where each agent estimates their individual pseudogradient via zeroth-order information, namely, measurements cost function values, as typical extremum seeking control. Third, generalize our setup results for multi-agent systems nonlinear dynamics. Finally, apply methods connectivity control robotic sensor networks almost-decentralized wind farm optimization.