Research On Observer-Based Security State Estimation For Cyber-Physical System

With the development of science and technology and the continuous improvement of people's demands for material life,our research on the security of cyber physical system is becoming more and more important.If a control system is attacked,the state estimation of the system will become inaccurate,which will affect the reliable control of the system,and can lead to system performance degradation or even system instability resulting in catastrophic accidents.Therefore,in order to ensure the safety and reliability of cyber physical system under attacks,it is very important to design an effective observer to estimate the state of the system.In the framework of linear cyber physical systems and lipschitz nonlinear cyber physical systems,this paper mainly studies the problem of safety state estimation.In order to improve the accuracy of system state estimation,three state observers are designed for effective state estimation based on different types of attacks.The research background and research status is introduced in the first chapter.Some lemmas and definitions are presented in the second chapter,which will be used in this article.In the third chapter,an adaptive sliding mode observer is designed for a class of linear cyber physical systems.Firstly,in the case of simultaneous attacks on system actuators and sensors,we assume that the attack has an upper bound,but we can not fully know the upper bound.We only know part of the information of the upper bound,and then we estimate the information of the unknown upper bound with an adaptive method.Then,combining with the idea of sliding mode observer,we propose a theorem for the existence of an adaptive sliding mode observer.By solving two linear matrix inequalities simultaneously,we can obtain the observer gain that needs to be designed.Thus,we can design our adaptive sliding mode observer for actuator and sensor attacks.Finally,we verify the validity of the observer proposed in this chapter by simulation.In the fourth chapter,for linear cyber physics systems and lipschitz nonlinear cyber physics systems,we design their new unknown input observers respectively.Firstly,we study attacks on the actuator of a linear system.Because the classical unknown input observer requires the unknown input matrix to be full rank,we improve the classical unknown input observer and propose a new unknown input observer for the actuator attack.By solving a linear matrix inequality,the matrix coefficients of the observer are designed to decouple a part of the actuator attack vectors so that they do not affect the state estimation.Then we use the H?performance to weaken the impact of the remaining part of the actuator attacks on state estimation.Secondly,we study the attacking situation of sensor in linear system.By using the principle of low-pass filter to extend the original system,we ingeniously convert the attacking of sensor to the actuator to a certain extent,and then design a new augmented unknown input observer for augmented system,we can make use of the conclusion of the above executor attack to do state estimation.Finally,we extend the new unknown input observer to lipschitz nonlinear system,and design a new type of unknown input observer when both the actuator and the sensor are attacked at the same time.By solving a linear matrix inequality,we also design the matrix coefficients of the observer.After decoupling one part of the actuator attack,the other part of the actuator attack,the sensor attack and the sensor attack derivative are expanded into an attack vector,which is optimized by using the H? performance.The validity of our research is proved by simulation experiments.In the fifth chapter,we address a class of lipschitz nonlinear cyber physical systems with gaussian white noises and unknown input disturbances.When both sensor and actuator are attacked,the state of the original system and the attacked signal of the sensor and the gaussian white noise of the output are augmented into the state of the augmented system by augmented method,and a simultaneous state observer is designed for the augmented system,that is,the state of the original system and other signals can be observed simultaneously.By solving a linear matrix inequality,the gain of the simultaneous state observer is obtained,and the unknown input disturbances and the actuator attack signal are weakened by H? performance.So,it has little effect on the state of the augmented system to be estimated.After the design of simultaneous state observer is completed,the state of augmented system can be estimated,and the state of original system,sensor attack signal and gaussian white noises can be estimated by certain operation.Finally,the validity of the simultaneous state observer designed in this chapter is verified by simulation.At last,conclusions and further research directions are given in the sixth chapter.