Chanhwa Lee

Secure and robust state estimation under sensor attacks, measurement noises, and process disturbances: Observer-based combinatorial approach

This paper presents a secure and robust state estimation scheme for continuous-time linear dynamical systems. The method is secure in that it correctly estimates the states under sensor attacks by exploiting sensing redundancy, and it is robust in that it guarantees a bounded estimation error despite measurement noises and process disturbances. In this method, an individual Luenberger observer (of possibly smaller size) is designed from each sensor. Then, the state estimates from each of the observers are combined through a scheme motivated by error correction techniques, which results in estimation resiliency against sensor attacks under a mild condition on the system observability. Moreover, in the state estimates combining stage, our method reduces the search space of a minimization problem to a finite set, which substantially reduces the required computational effort.

On robust stability of disturbance observer for sampled-data systems under fast sampling: An almost necessary and sufficient condition

Despite increased interest on discrete-time disturbance observer (DOB) in both theory and application, study of robust stabilization via the DOB does not seem to be mature. This is because most of the existing studies on robust stability are based on the small-gain theorem, so that the results are just sufficient conditions and the amount of uncertainty that the DOB-based control system can tolerate is conservative. In this paper, motivated by a recent work on the continuous-time case, we present an almost necessary and sufficient condition for robust stability of the sampled-data system controlled by a discrete-time DOB under fast sampling. In particular, our study clarifies the phenomenon that the sampling process can hamper stability of the DOB-controlled systems by generating additional (and possibly nonminimum-phase) zeros, and explains why the blind discretization of the continuous-time DOB controller using fast sampling may fail. For the robust stabilization against both these extra zeros and plant uncertainty, we also propose a new design methodology for the nominal model and the Q-filter. It turns out that arbitrarily large uncertainty can be compensated by appropriately designing the Q-filters and by fast sampling. A benchmark problem is revisited to illustrate the validity of the proposed analysis and design method.

When adversary encounters uncertain cyber-physical systems: Robust zero-dynamics attack with disclosure resources

In this paper we address the problem of designing a robust stealthy attack for adversaries to compromise an uncertain cyber-physical system without being detected. We first re-interpret the zero-dynamics attack based on the normal form representation. Then, a new alternative zero dynamics attack is presented for uncertain systems. This alternative employs a disturbance observer and does not require exact system knowledge in order to remain stealthy. The proposed robust zero-dynamics attack needs a nominal model of the system and, in addition, utilizes the input and output signals of the system. The proposed attack illustrates how the adversary is able to use disclosure resources instead of exact model knowledge. A simulation result with a hydro-turbine power system is presented to verify the attack performance.