Research On Nonlinear Filtering Methods For State Estimation Of Cyber-physical Systems

Nonlinear state estimation plays an important role in cyber-physical systems(CPS).However,The observations of nonlinear filter,subject to the transmission network of CPS,greatly suffer from data packet loss,channel fading and limited channel capacity.By fully considering the influence of the transmission network,this paper address the practical design and the stochastic stability problems of nonlinear filter with data packet loss and channel fading.Also,the nonlinear event-trigger filter are developed to overcome the limited channel capacity and get relieved of communication burden.In order to tackle the data packet loss,Chapter 2 addresses the stochastic stability problem of the extended Kalman filter(EKF)with intermittent observations.With transmitted observation of the filter modeled as a Bernoulli process,the existence of a crucial arrival rate is proved such that the prediction error covariance matrix(PECM)is mean bounded when the arrival rate exceeds a threshold value.Moreover,the sufficient conditions is given regarding the stochastic stability of estimation error,which requires the mean boundedness of PECM,appropriate system constraints and small enough initial estimation error.In view of the practical application suffering from both data packet loss and channel fading,Chapter 3 further proposes a modified extended Kalman filter(MEKF).Bernoulli process and Rayleigh fading are taken into consideration to model data packet loss and channel fading,respectively.Sufficient conditions are established for the boundedness of the expectations of the PECMS of MEKF,which shows the existence of a crucial arrival rate.The upper bound of PECMS is also derived by the property of exponential integral.Further sufficient conditions are provided for mean-square bounded estimate error of the MEKF using the fixed-point theorem.Furthermore,in order to deal with the limited channel capacity,an event-trigger extended Kalman filter(ET-EKF)is established in Chapter 4.An event-trigger strategy is proposed to guarantee that only the observations containing innovational information are transmitted.The corresponding filtering algorithm is also proposed to utilize the information in event-trigger strategy.Finally,its feasibility and performance is demonstrated using the standard IEEE 39-bus system.To satisfy the invariable arrival rate need of the limited capacity channel,an eventtrigger particle filter is accordingly designed in Chapter 5.An arrival rate guaranteed event-trigger strategy is established by utilizing Monte Carlo methods to approximate the prior condition distribution of observation.Moreover,an ET-PF filtering algorithm is further proposed by making full use of the information from the event-trigger strategy,to enhance the performance of estimation.IEEE 39-bus system are also utilized to verify the feasibility of the designed ET-PF.Under the constraints including both communication and computation power at CPS nodes,an event-trigger heterogeneous nonlinear Kalman filter(ET-HNF)is designed in Chapter 6.ET-HNF utilizes the unified filtering of unscented transformation with PF theories so that the accuracy and the relief of communication burden can be both guaranteed.An unscented transformation based event-triggered UKF(ET-UKF)is firstly designed to supply the event-triggered strategy.Furthermore,a Monte Carlo based filtering algorithm is designed in the estimation center to provide accurate filtering results.The feasibility and performance of ETH-NNF is verified using the standard IEEE 39-bus system.