Research On Event-triggered State Estimation With Unknown Input

In recent years,networked state estimation is widely used in intelligent transportation,Internet of Things and other fields because of its low cost,long-distance transmission and high reliability.However,Networked state estimation is faced with the problems of limited communication bandwidth,loss of measurement,and external disturbance.Due to limited energy of sensors,aging of devices and network congestion,high energy consumption,missing and intermitten of measurement data are easily occured in the process of data transmission,which leads to the degradation of state estimation performance.Most of the existing network state estimation methods are focused on Gaussian noise,but in the actual system,due to the measurement outliers,modeling errors and other reasons,the noise presents non-Gaussian heavy-tailed characteristics.Thus,the measurement modeling method based on Gaussian distribution is not accurate.In addition,the sensor network always faces the disturbance or attack from the internal and external systems,which leads to the degradation of estimation accuracy and even divergence.In this paper,we focus on solving state estimation,unknown input high energy consumption,the networked state estimation problem.The main results are as follows:(1)For a class of non-Gaussian nonlinear systems with heavy-tailed noise,the problem of state estimation with limited energy and missing measurement in the process of data transmission is considered.Firstly,the Student’s t distribution is used to model heavy-tailed non-Gaussian noise,and a binary Bernoulli distribution is introduced to represent the loss of measurement data.Then,the measurement-based event triggering mechanism is designed to reduce the number of network transmission and save the communication bandwidth.Under the event-triggered mechanism,the Student’s t filter is designed to obtain the state estimation of the nonlinear system.At the same time,some sufficient conditions for the boundedness and stability of the estimation error under the packet loss rate constraint are given.Simulation examples show the effectiveness of the proposed method.(2)For a class of heavy-tailed non-Gaussian linear stochastic systems with unknown inputs,the state estimation problem is studied.A three-step iterative Student’s t filter is proposed.Under the assumption of full rank of disturbance input gain matrix,the estimation of unknown input is obtained in terms of the principle of unbiased estimation and least squares sum of estimation error.The state estimation is obtained by combining event-triggered mechanism with Student’s t filter.The sufficient conditions for the estimation error system to be stable are given.The simulation example verifies the effectiveness of the method and compares with the existing methods.(3)For a class of non-Gaussian linear stochastic systems with unknown input and heavy-tailed noise,the problem of state estimation with communication constraints and missing measurement is studied.Firstly,the missing measurement model is established.Secondly,the measurement-based triggered mechanism is intrudeced to reduce the comunication burden.Finnaly,the filter is designed to estimate unknown input and system state simultaneously.The simulation results show that the more times the measurement data are lost,the lower the estimation accuracy;it shows that the estimation accuracy of unknown input and system state obtained by the proposed method is higher than the existing methods when the noise is heavy-tailed.