Decentralized Optimization Algorithm of Networked Systems

Zhe Liu,Xi Chen,Xiaohong Guan,Li Zhao
DOI: https://doi.org/10.1109/chicc.2016.7553797
2016-01-01
Abstract:We consider a large-scale networked system which consists of multiple subsystems. With decentralized optimization algorithms, each subsystem can cooperate with other ones to obtain the optimal solution to an optimization problem. This paper presents a decentralized optimization algorithm based on penalty functions. Rather than most existing decentralized optimization algorithms, this algorithm does not require convexity of optimization problem. Firstly, the coupling between subsystems is reduced by variable splitting method. Then, the penalty function is introduced to weaken the restriction on subsystems. With knowledge of its neighboring subsystems, each subsystem can obtain the local/global optimal solution of the original optimization problem. Illustrative examples are presented for demonstration.
What problem does this paper attempt to address?