The complexity of managing multidisciplinary engineering systems offers an unprecedented opportunity to investigate decomposition methods, which separate a system into a number of smaller subsystems that can be designed in multiple physical locations and coordinate the design of the subsystems to collaboratively achieve the original system design. This paper studies a network target coordination model for optimizing subsystems that are distributed as multiple agents in a network. To solve these coupled subsystems concurrently, we consider the “consensus optimization” approach by incorporating subgradient algorithms so that the master problem or auxiliary design variables required by most distributed coordination methods are not needed. The method allows each agent to conduct its optimization by locally solving for coupling variables with the information obtained from other agents in the network in an iteratively improving process. The convergence results of a geometric programming problem that satisfies the convexity assumption is provided. Moreover, two non-convex examples are tested to investigate the convergence characteristics of the proposed methods.
