Distributed Nonconvex Optimal Resource Allocation via a Momentum-Based Multiagent Optimization Approach

Abstract

In this article, a momentum-based multiagent optimization approach is developed for distributed nonconvex optimal resource allocation. The proposed resource allocation model is formulated without the convex conditions, and a paradigmatic system based on the gradient descent with momentum method is proposed for handling its functional nonconvexity. Based on the paradigmatic system, a momentum-based multiagent system (MAS) is developed, and its convergence and convergence rate to a local minimizer are proven. Then, a distributed average tracking approach is introduced, based on which a hybrid multiagent optimization approach consisting of multiple MASs and a meta-heuristic rule is designed for seeking global minimizers. Finally, a simulation in a chiller system is elaborated to demonstrate the enhanced stability, fast convergence, and optimality of the developed distributed optimization approaches.