A Scaling-Function Approach for Distributed Constrained Optimization in Unbalanced Multiagent Networks

Abstract

This article aims at developing a scaling-function approach for distributed optimization of unbalanced multiagent networks under convex constraints. The distinguishing feature of the algorithm is that it does not employ agents’ out-degree information, nor does it require the estimation of the left eigenvector, corresponding to the zero eigenvalue, of the Laplacian matrix. Existing approaches for unbalanced networks either demand the knowledge on agents’ out-degrees, which is impractical in applications, where an agent might not be aware of the detection and employment of its information by other agents, or require every agent to be equipped with a network-sized estimator, causing an additional <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$n^2$</tex-math></inline-formula> storage and communication cost with <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$n$</tex-math></inline-formula> being the network size. The results exhibit an inherent connection between the selection of the scaling factor and the convergence property of the algorithm, among other known factors such as the network topology and the boundedness of the subgradients of the local objective functions. Numerical examples are provided to validate the theoretical findings.