Secure Distributed Kalman Fusion Estimation For Cyber-Physical Systems Under FDI Attacks

As one of the important issues in cyber-physical systems(CPSs),distributed fusion estimation has become a major concern because of its fast computing speed,high reliability and strong fault tolerance,and has been intensively applied in industrial control systems,smart home and other fields.However,the networked distributed structure makes it vulnerable to cyber security threats.Meanwhile,the linearization errors,which increase the design difficulty of the nonlinear fusion estimation system,bring new challenge to the analysis and design process under cyber attacks.Therefore,it is of great theoretical significance and practical relevance to study the secure distributed fusion estimation problem for the CPSs.False data injection(FDI) attacks,which are to destroy the integrity of CPSs information by tampering with the measurement and control data,can greatly damage the performance of system fusion estimation.As one of the typical cyber attacks with high level of threat to CPSs,it is stealthy,intrusive and goal-oriented.Based on the distributed information fusion theory,this thesis proposes new nonlinear secure state fusion estimation methods,which focuses on solving the problem of FDI attacks in nonlinear CPSs.The main works and results are as follows:1)The secure distributed fusion estimation problem for a class of nonlinear CPSs is investigated,and the characteristics of typical cyber attacks are introduced in detail,which include attack strategies,working modes and damage to the systems.The influence of linearization errors on the performance of distributed fusion estimation is also analyzed.Therefore,this thesis mainly studies two kinds of problems,namely,secure distributed fusion estimation for nonlinear CPSs when the sensors are attacked or the actuators are attacked.2)The secure state fusion estimation problem for a class of nonlinear CPSs whose sensors are attacked by FDI is investigated,and the attacked data is eliminated by the chi-square detector based on the Kalman filtering method.To estimate the system state with incomplete measurements,a progressive cubature information filtering(PCIF)method is designed based on progressive gaussian filtering and cubature integration criterion.Then,the modified PCIF method is employed as local estimators for distributed fusion systems,thus the distributed fusion results of linear systems are extended to nonlinear systems.Finally,simulations of a mobile robot tracking example are presented to demonstrate the advantage and effectiveness of the proposed methods.3)The secure state fusion estimation problem for a class of nonlinear CPSs whose actuators are attacked by FDI is investigated,and the attack signal is modeled as a time-varying parameter in the system state equation.To realize the joint estimation of system state and attack signal,an adaptive extended Kalman filtering(AEKF)is designed as local estimators based on the extended Kalman filtering method and the recursive least square method.Then,by introducing the compensation factor into the fusion center,a nonlinear distributed weighted fusion estimation method is proposed to improve the performance of the fusion estimation.Finally,the feasibility and effectiveness of the proposed method are verified by the numerical simulation and mobile robot target tracking simulation.