Accelerated Distributed Gradient Tracking for Constrained Aggregative Optimization Over Time-Varying Digraphs

This article investigates the distributed, constrained aggregative optimization problem over time-varying digraphs. All agents in the multiagent network aim at collaboratively minimizing a sum of objective functions depending jointly on the local decision variables and an aggregative variable of all involved agents under local, closed convex set constraints. A distributed algorithm based on the Push-DIGing framework is proposed to resolve the aggregative optimization problem. In addition, the method of feasible directions is embedded into the proposed algorithm to address the involved set constraints. Furthermore, the momentum-based accelerated techniques are incorporated into the proposed algorithm to obtain better convergence performance. Under the strongly convex properties of objective functions, it is rigorously established that a linear convergence rate is achieved by the proposed algorithm. Finally, a simulation example of the proposed algorithm in the multirobot surveillance problem is provided to verify its effectiveness.