Controllability of Networked Systems

Controllability is an important prerequisite of effective system control. Due to the large scale, complex structure, multi-dimensional dynamics, and various sampling patterns, classical controllability theories are confronted with application problems in networked systems. Recently, new controllability analysis techniques with sufficient and/or necessary conditions have been developed for both continuous-time LTI networked systems and corresponding sampled-data systems. Results reveal that for networked systems with periodic sampling on both control and transmission channels, the pathological sampling of single node systems can be eliminated by the joint effects of topologies and inner-couplings. However, for systems with singular topology matrix (such as star and chain-structure), the pathological sampling of node systems will definitely cause the whole system to lose controllability. Easier-to-verify criteria are developed for systems with nonidentical dynamics and multilayer network structures, and the computational complexity is decreased compared to the existing results. Moreover, typical multi-rate sampling patterns are investigated, where sampling periods are distinct on different types of channels.