Distributed Optimization Algorithms Based on Multi-Agent Networks

Distributed optimization has many applications in such fields as cooperative control, information processing and game theory. Although several distributed algorithms have been developed for some convex optimization problems in the literature, how to efficiently solve large-scale constrained optimization problems still remains a challenging problem and appeals for more efficient alternatives. This project aims to develop distributed algorithms for cooperative optimization among multiple agents connected through a network. The goal is to optimize global objective functions which are combinations of local objective functions known by the agents only. We will focus on modeling and consensus analysis of multi-agent networks for constrained optimization problems. The Lagrangian decomposition and penalty function methods will be investigated for the design of algorithm. More efficient consensus criteria, which lead to distributed algorithms for optimization problems, will be developed for multi-agent networks. Numerical comparisons with existing optimization algorithms will be carried out. The ultimately expected outcome of the project will be more efficient distributed algorithms for constrained optimization problems.