Distributed Optimization for Time-Varying Multi-agent Networks Under Partial Information: A Game Theoretic Approach

In this paper, the distributed optimization problem of time-varying multi-agent networks is investigated by using game approaches. First, a game model is used to describe the distributed optimization problem. Then, the optimization objective can be achieved at the corresponding equilibrium of the game. This design method has the advantage of robustness to many uncertain factors; Even if the communication is delayed and asynchronous, the resulting equilibrium can solve the distributed optimization problem. Furthermore, a discrete-time distributed algorithm, called the utility-based distributed projection subgradient (UDPS) algorithm, which guarantees convergence, is devised to search the Nash equilibrium (NE) of the formulated game. Finally, numerical simulations prove the previous conclusions.