Distributed Optimization With Coupling Constraints

Abstract

In this article, we investigate distributed convex optimization with both inequality and equality constraints, where the objective function can be a general nonsmooth convex function and all the constraints can be both sparsely and densely coupling. By strategically integrating ideas from primal-dual, proximal, and virtual-queue optimization methods, we develop a novel distributed algorithm, referred to as IPLUX, to address the problem over a connected, undirected graph. We show that IPLUX achieves an <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$O(1/k)$</tex-math></inline-formula> rate of convergence in terms of optimality and feasibility, which is stronger than the convergence results of the alternative methods and eliminates the standard assumption on the compactness of the feasible region. Finally, IPLUX exhibits faster convergence and higher efficiency than several state-of-the-art methods in the simulation.