Research On Distributed Kalman Filter Based On Maximum Correntropy Criterion

Multi-sensor network is a new sensor deployment model that integrates multiple technologies.It consists of a large number of sensors with computing capabilities deployed in a certain range.It can decentralize and cooperate with each other to complete specified tasks Stability and scalability.One of the most basic applications of multi-sensor networks is distributed estimation or tracking.As a widely used data filtering technology with multiple extension methods,Kalman filtering is distributed by researchers to transform it into distributed Kalman filtering,which can be applied to multi-sensors by exchanging information with neighboring nodes in a networked distributed environment.The traditional Kalman filter is an optimal filter defined in the Gaussian distribution.It has a very poor effect on non-Gaussian noise,especially impulse noise.The distributed Kalman filter is only a simple distributed extended version,so it also has this problem.At the same time,non-Gaussian noise such as impulse noise appears very common in many practical scenarios,such as target tracking and digital communication,so there has been a lot of research on making Kalman filtering work in this noise environment.In recent years,a maximum correntropy Kalman filter proposed to deal with this question.This algorithm performs better than traditional Kalman filters in non-Gaussian environments,especially under impulsive noise.In this paper,the distributed maximum correntropy entropy Kalman filter is derived by distributing the maximum correntropy Kalman filter.This distributed Kalman filter can effectively deal with non-Gaussian noise in a distributed environment.Finding a suitable consensus strategy to make the system work together while maintaining the consistency of the information is a basic problem of distributed estimation.The autonomous cooperation between multi-sensor networks requires meeting high accuracy and consistency,however,each sensor is located differently,the network topology is different,and the interference situation is not the same,the distributed estimation algorithm is often different within a certain range on each sensor.Because distributed Kalman can only interact with neighboring nodes,the information of one sensor is still local information,so special processing needs to be done on the algorithm so that the local information can be integrated with other local information as quickly as possible and reach an agreement.This is the role of the consensus algorithm.This paper applies consensus theory to the distributed maximum correntropy Kalman filter.First,several existing traditional consensus strategies are introduced into the distributed maximum correntropy Kalman filter,and then a new method based on the maximum correntropy cost function is proposed.Then theory analysis has proved the effectiveness of the above consensus method.Simulation results show that the algorithm effectively reduces the state difference between nodes.