Unknown Input Observer Based Resilient Control System Design

Abstract

Resiliency of control systems means characteristics that a system operates well under malicious attacks on instrumentation. In this thesis, we propose Unknown Input Observer (UIO) based resilient control system design method for the case of simultaneous attacks on sensors and actuators. We use an observer based Resilient State Estimation (RSE) method for protecting from sensor attack. Limitation of existing RSE method is that state estimation error is affected by disturbance and actuator attack. In order to supplement the drawback, we propose the UIO based RSE method. Proposed method provides superior state estimation performance regardless of the effect of disturbance and actuator attack. In addition, we suggest an algorithm that would be able to diagnose which sensors are attacked or broken. In order to protect the system from actuator attack, we propose a method that would be able to estimate and reject the effect of actuator attack. Proposed method exploits UIO based RSE method. The proposed method provides a level of resiliency to the control system when disturbance and attack are slow varying. Simulation and experiment are implemented on the magnetic levitation system platform so as to validate the effectiveness of proposed method. In addition, we compare the performance between existing method and proposed method. ⓒ 2017 DGIST