Alternating direction method of multipliers in optimal control of systems of systems

Abstract

This paper investigates the usage of a distributed algorithm — the alternative direction method of multipliers — for control of system of systems, i.e. control of a complex system comprising large number of distributed and partially autonomous subsystems. The control algorithm is based on parametrization and decomposition of the central optimization problem into local optimization problems specific to individual subsystems, while taking into account the coupling constraints that link them. Local problems are solved offline using multi-parametric optimization methods, whereas the coordination problem is solved online at every sampling step and the optimal solution is applied to the system at hand in a receding horizon fashion. The method solves the optimal control problem online iteratively while exploiting the control system structure. The efficiency of the method is tested on a number of generated distributed optimization problems modeling large systems of systems.