Asynchronous Distributed Optimization for Network Utility Maximization Problems with Event Triggered Communication

Abstract

This paper is concerned with an event triggered distributed optimization algorithm for network utility maximization (NUM) problems. According to an event triggering logic, a source broadcasts its information to links when a local signal exceeds a state dependent threshold. A similar communication logic is executed by each link, where the link broadcasts its information to all sources that use the link. The algorithm is based on a sequential barrier method, which can be applied to optimization problems with constraints. The efficiency of the proposed scheme is verified via simulations. The simulation result shows that the proposed algorithm reduces the number of message exchanges while guaranteeing the converge to the optimal solution.