Methods of state estimation resilient against sensor attacks and robust against exogenous disturbances

Abstract

Control systems that are resilient against malicious attacks have recently become one of the important research topics. This is due to the fact that many critical infrastructures in our society such as power grid, nuclear facility, and public transportation, rely on feedback control technology, and that malfunction of said systems due to attack could be disastrous. Malicious attacks on control systems indeed occurred in recent years. The problem of resilient state estimation against sensor attacks, which is concerned of correctly estimating plant state despite of sensor attacks, have been previously addressed. Existing methods, however, lose the accuracy of the estimation if external disturbance exists. In this paper, we investigate the problem of resilient state estimation that are robust against external disturbance. Specifically, unknown input observer based state estimation has been developed to eliminate the effect of external disturbance on the state estimation that are resilient against malicious attacks on sensors. Associated design conditions have been derived. Additionally, disturbance observer based state estimation method has been developed. Validation of the proposed methods is provided through experiments.