Research On Underwater Integrated Navigation Method Based On Robust Adaptive Filtering Under Maximum Correntropy Criterion

Through autonomous underwater vehicles(AUV),human beings can develop marine resources and study underwater ecological environment more easily and safely.In order to ensure the success of AUV activities in both military and civilian fields,the requirements for the navigation and positioning accuracy of AUV are gradually improving.Integrated navigation has always been the main way of navigation and positioning for AUV.Because the underwater environment is complex and changeable,the measurement noise of the underwater integrated navigation system will have outliers,and the accuracy of the navigation system will be affected.In this paper,a new robust adaptive filtering algorithm is proposed,and an underwater integrated navigation method based on this algorithm is designed after full considerate the complexity of underwater conditions.Firstly,the underwater integrated navigation system has been introduced.The underwater integrated navigation system includes strapdown inertial navigation system(SINS)and Doppler log.For SINS and Doppler Log,their basic principles are introduced.Then the direct and indirect filtering methods of integrated navigation system are introduced.Then,the filtering algorithm used in integrated navigation is studied.The classical Kalman filter has been introduced.In view of the non-Gaussian noise distribution caused by outliers,robust filtering is studied.Kalman filter based on Huber and maximum entropy Kalman filter(MCKF)are introduced.Through the simulation experiment of position tracking model under the presence of outliers in process noise and measurement noise,the filtering algorithms introduced in this chapter are compared and studied.Then,in view of the fact that the noise can not be estimated continuously and steadily due to the complex underwater environment,so that the accurate process noise covariance matrix(Q)and measurement noise covariance matrix(R)can not be obtained,Sage-Husa adaptive filtering and Variational Bayes Adaptive Kalman Filter(VBAKF)are studied.Through the simulation experiment of bearings-only tracking model with slowly varying Q and R,the two adaptive filtering methods studied in this section are compared and analyzed.Finally,in view of the inaccuracy of Q and R in the filtering model of integrated navigation system and the impact of impulse noise,Maximum Correntropy Variational Bayes Adaptive Kalman Filter(MCVBAKF)is proposed.MCVBAKF takes the result of maximum entropy convergence as a priori condition,and uses the variational Bayesian method to update the posterior.It has strong robustness and adaptability.MCVBAKF can not only deal with impulse noise effectively,but also adaptively correct the influence of inaccurate Q and R.Finally,an underwater integrated navigation method based on MCVBAKF is designed and validated in SINS/DVL integrated navigation model.The simulation results show that the proposed underwater integrated navigation method based on MCVBAKF has higher estimation accuracy than underwater integrated navigation method based on KF and underwater integrated navigation method based on MCC when the Q and R in the integrated navigation system is inaccurate and the system is subjectet to impulse noise.