Distributed Extremum Seeking for Constrained Convex Optimization With Input Dead‐Zone

This paper studies a distributed constraint optimization problem for a multi‐agent system with unknown cost functions and asymmetric input dead‐zone. First, the constraint optimization problem is transformed into an unconstrained optimization problem by using the exact penalty method. As only measurements of the cost functions and local inequality constraints are available, a distributed extremum‐seeking algorithm is presented based on an event‐triggered communication scheme to reduce communication frequency. Moreover, to address the impact of input dead‐zones, a dynamic compensation scheme is introduced into the designed algorithm. Due to the dual time‐scale structure of the designed distributed algorithm, singular perturbation techniques are used to achieve semiglobally practically asymptotical convergence of the decision variables of all agents to the optimal solution. Finally, simulations verify the effectiveness of the proposed algorithm.