Target Controllability of Structured Systems

This paper deals with controllability of structured linear time-invariant systems. Controllability of large complex systems are often difficult to attain as a substantial part of nodes in the network require to be actuated and the amount of energy required is large. Moreover, often it is unnecessary to control the entire dynamics of the network: instead, only certain portions of the network need to be controlled. This concept is known as target control. In this paper, we address target control for large complex systems using their network topology. We prove that if there exists one numerical system in the family of the network that is target controllable, then almost all systems in the family are target controllable. This result thus concludes that target controllability is a generic property. Then we propose a bipartite-matching based condition to determine target controllability of a subset of nodes in a single input network based on the generic rank of the controllable subspace. Finally, we provide experimental analysis to complement our results using real-world data sets.