Distributed Optimization for Networked Multi-Agent Systems

Background

Recently, applied optimization has had a tremendous impact on the development of algorithms for coordination, control, and resource sharing in networks and systems. Examples include multi-robot coordination, distributed estimation, radio resource management in wireless systems, and traffic engineering in data networks. As the size of these networks grows along with the level of their required performance, the interaction between subsystems and the role of system constraints cannot be neglected any further. Striving for optimal operation using distributed controllers and limited, local information processing provides practical motivations for developing tractable distributed control and optimization methods for large-scale systems. The reason why centralized solutions are avoided in many contexts is that gathering data in a central location and performing operations there is inflexible and costly, or even prohibitive when the number of the subsystems grows very large (hundreds of thousands to millions). In addition, distributed algorithms and the distribution of computations can mitigate complexity issues by solving smaller and simpler optimization problems instead of the global one.

Project team members

  • Arman Sharifi Kolarijani
  • dr. Peyman Mohajerin Esfahani
  • dr.ir. Tamas Keviczky

Objective

The main goal of the project is to investigate and develop distributed optimization and decision-making algorithms for networked, large-scale multi-agent systems. We are especially interested in applications involving real-time control and estimation problems for interconnected dynamical systems. Potential research topics include consensus optimization, ADMM, distributed event-triggered control, and distributed approaches to mixed-integer optimization problems.