Examination Committee

Prof Chih-Chen CHANG, CIVL/HKUST (Chairperson)
Prof Ling SHI, ECE/HKUST (Thesis Supervisor)
Prof Yang SHI, Department of Mechanical Engineering, University of Victoria (External Examiner)
Prof Shaojie SHEN, ECE/HKUST
Prof Wai Ho MOW, ECE/HKUST
Prof Francesco CIUCCI, MAE/HKUST

Abstract

Constrained remote state estimation and quickest change detection in the context of Cyber-physical systems are studied in this thesis. The remote state estimation is studied from the perspective of both an estimator and an attacker. Three different problems are formulated, and the common theme is twofold: (1) The resources for the estimator/attacker/decision maker are limited, and efficient resources utilization policies are thus necessary. (2) In each proposed (optimal) policy, decisions are made sequentially with real-time information.

In the problem of remote estimation from an estimator's perspective, a wireless fading channel is adopted. The successful transmission probability depends on both the channel gains and the transmission power used by the sensor. Jointly optimal transmission power controller and remote estimator are presented by formulating the problem as a partially observable Markov decision process. Structural results are further provided using the majorization theory. The remote estimation from an attacker's perspective is studied in multi-systems. An attacker is present and may generate noises to exacerbate at most N of the M communication channels between sensors and the fusion center at each time. The optimal attack policy is solved as a solution to a Markov decision process, of which a threshold structure is proved. We further provide an asymptotically optimal (as M and N go to infinity) policy, which is quite easy to compute and implement. Unlike the classical quickest change detection problem, in this thesis work the decision maker chooses one of two different sequences of observations about the monitored target at each time. The information quality and sampling cost of these two sequences are different. We present an asymptotically optimal joint design of observation scheduling policy and stopping time such that the detection delay is minimized subject to constraints on both average run length to false alarm and average cost per sample.