Examination Committee

Prof Weichuan YU, ECE/HKUST (Chairperson)
Prof Li QIU, ECE/HKUST (Thesis Supervisor)
Prof Ling SHI, ECE/HKUST

Abstract

Networked control systems (NCSs) are the control mechanism with noisy communication channels in between, which are ubiquitous in modern computing and guided devices and systems. Stabilization plays the most respected and dominated role in NCSs. While there are lots of research work on single input systems, how to describe the quality of MIMO channels, how to design the controller and the transceiver jointly, and how to characterize fundamental performance limits in control setting with both MIMO plants and MIMO channels are open to study.

In the first part, we will study those issues in the context of additive white Gaussian channels. First of all, we will use the concept of channel capacity from classic information theory to quantify the channel, where a necessary condition and a sufficient one are derived, respectively. Generally, there exists a gap between these two due to the closed-loop characteristics of the NCSs. Then the incremental mutual information has been introduced and defined, in term of which a necessary and sufficient condition has been given. Finally, how to design the optimal controller and the transceiver with the gap minimized is investigated.

In the second part, the multiplicative noise channel has been studied. Firstly, we have researched the issues about how to use positive system techniques to analyze the system variances, how to design the control and communication scheme, and what conditions of stabilization are. Secondly, the results are further extended to a generalized stable plant with the transceiver as the only design freedom. Lastly, the poles and zeros together with their associated directions have been used to study the output feedback case.