Maximum Correntropy Based Particle Filter And Its Application In Underwater Cooperative Navigation

Autonomous underwater vehicle(AUV)is a powerful assistant for human to explore the ocean.With the development of human exploration of the ocean,the single AUV has been unable to meet the requirements.Multi-AUV cooperative navigation system can work together and has great advantages compared with single AUV.In cooperative navigation,the position of AUV needs to be estimated by using the non-linear filter based on state space model.The commonly used non-linear filter include extended Kalman filter(EKF),unscented Kalman filter(UKF),cabuture Kalman filter(CKF),particle filter(PF)and so on.However,due to the complex marine environment,the measurement noise of multi-AUV cooperative navigation system often appears outliers.At this time,the estimation accuracy will be reduced if commonly used non-linear filtering method is used.Therefore,it is necessary to find a more appropriate filtering method.In this paper,based on particle filter,multi-AUV cooperative navigation technology with outliers in measurement noise is studied in depth.Firstly,this paper introduces the development status of multi-AUV cooperative navigation,multi-AUV cooperative navigation filtering algorithm and particle filter at home and abroad,and summarizes the commonly used navigation methods based on AUV,then elaborates the basic principle and mathematical model of multi-AUV cooperative navigation.Then,the problem of reducing the filtering accuracy of traditional PF in the presence of outliers in measurement noise is studied.Based on Bayesian theory,PF and three improved PFs are introduced and deduced in detail.Then,the maximum entropy criterion(MCC)and the maximum entropy unscented Kalman filter(MCUKF)are introduced.Based on this,the maximum entropy unscented particle filter(MCUPF)is proposed.This method is based on the unscented particle filter,and is improved by using the maximum entropy criterion,so that it has robustness in the presence of outliers in measurement noise and can maintain high filtering accuracy.Finally,the effectiveness of the proposed filtering method is verified by simulation experiments.Then,the problem of high computational complexity of MCUPF proposed in the previous chapter is studied.Firstly,three common PF resampling methods are introduced,and then a KLD resampling method which can adjust the number of particles in real time is introduced.Based on the MCUPF proposed in Chapter 3,considering the problem of excessive computation,the traditional resampling method is replaced by KLD resampling method,which can adjust the number of particles adaptively.A maximum entropy adaptive unscented particle filter(MCAUPF)is proposed.Finally,the effectiveness of the proposed filtering method is verified by simulation experiments.Finally,a multi-AUV cooperative navigation method based on MCAUPF is proposed to solve the problem that the performance of the existing cooperative navigation method based on improved particle filter will be greatly degraded in the presence of measurement noise outliers.The simulation results show that the proposed filtering method has higher estimation accuracy and lower computational complexity compared with the existing robust particle filter and robust filtering methods in multi-AUV cooperative navigation system with outliers in measurement noise.