A decentralized coordination strategy for networked multiagent systems

We propose a decentralized coordination strategy for abstract networked multiagent systems in which individual agent goals are modeled as optima of privately known objective functions and the optimal network configuration as the optimizer of the sum of these functions. The strategy is based on a recently proposed decentralized optimization method, which we generalize here by allowing agents to update only a subset of optimization variables. We provide a Lyapunov argument for the global practical asymptotic stability of collective optimum for the proposed scheme.