Distributed Filtering For Stochastic Nonlinear Discrete Systems Over Sensor Networks

In recent years,the sensor networks,which are formed by a number of sensor nodes and have the capability of collecting information,processing information,and transmitting information,have been widely used in military management,environmental testing,medical and industrial production,and the like.With the rapid development of microelectronics technology and digital signal technology,distributed computing technology have used its advantages to accelerate the transformation of sensor networks in practical production applications,and have made many new applications for sensor networks.For the same monitoring area,different sensor nodes and the same sensor nodes at different locations have differences in the process of information perception and information processing,which makes the overall estimation accuracy decrease for the filtering algorithm.Therefore,the problem for improving the accuracy of sensor networks estimation is urgent to be solved.Based on the summary of previous research results,this paper studies the distributed filtering problem for stochastic nonlinear discrete systems over sensor networks by using recursive algorithm and consensus algorithm.The main research results are as follows:Firstly,for discrete time-varying systems with randomly occurring nonlinearities,consider two cases: consider designing the distributed filter by introducing MEF protocol in the communication process.Using the distributed filtering algorithm,the upper bounds of the one-step prediction error covariance and the filter error covariance are obtained by the variance constraint method.The upper bound of the filter error covariance is minimized to obtain an optimal filter gain.The feasibility and effectiveness of the proposed distributed filtering algorithm are verified by a numerical simulation.Further,considering the systems with parameter uncertainties and multiplicative noise,we design the consensus distributed filter,fuse the consistency algorithm and the recursive filtering algorithm,respectively.By the method of variance constraint,the one-step prediction error covariance and the filter error covariance can be acquired for each node.Then the upper bound for the filter error covariance of each node is minimized,and the optimal distributed filter gain and consensus gain for each node are obtained.The feasibility and effectiveness of the proposed consensus distributed filtering algorithm are verified by a numerical simulation.Secondly,for discrete time-varying systems with random uncertainties,consider two cases: consider the phenomenon of sensor missing measurements,and design the resilient distributed filter.Using the resilient distributed filtering algorithm,obtain the upper bound for filter error covariance,and then the upper bound of filter error covariance is minimized,finally a new matrix simplification technique is introduced to obtain the resilient filter gain.The feasibility and effectiveness of the proposed resilient distributed filtering algorithm are verified by a numerical simulation.Next,considering the existence of nonlinearities and sensor missing measurements for probability uncertainty systems,designing resilient consensus distributed filter,and using the resilient consensus distributed filtering algorithm to obtain the each node's upper bound for the filter error covariance.Furthermore,the resilient filter gain and the consensus gain for each node are obtained.The feasibility and effectiveness of the proposed resilient consensus distributed filtering algorithm are verified by a numerical simulation.