Research Of Underwater Terrain Aided Navigation Based On Multibeam Bathymetry Sonar

Underwater vehicle,as an important tool to exploit the ocean for human,has been widely applied in military and civilian fields,such as marine environment monitoring,submarine pipeline detection,theater reconnaissance and attack enemy ships or submarines.Accurate and autonomous navigation of underwater vehicle provide basic guarantee for long-term safe navigation and accurate detection.Underwater Terrain aided navigation,as a non-traditional navigation method,has good passivity,concealment,precision and autonomy because of its error does not accumulate with the time.In recent years,it has become one of the research hotspots in underwater navigation field at home and abroad.In this paper,underwater terrain aided navigation method using multi-beam sonar is carried out due to the ability of multi-beam sonar to measure two-dimensional terrain.some key technologies are studied,such as high-precision underwater digital terrain map,robust and fast terrain matching localization using multi-beam sonar and robust terrain tracking navigation based on recursive Bayesian estimation.The main work of the thesis are as follows:1.Studying the structure,mathematical model and theoretical method of underwater terrain aided navigation.Firstly,the components and mechanism of underwater terrain aided navigation system are analyzed,including underwater prior terrain database,reference navigation unit,terrain measurement unit and terrain matching unit.Then,the general mathematical model of underwater terrain aided navigation is summarized.According to the different coupling modes between reference navigation unit and terrain matching unit,the state models are divided into tightly and loosely coupled models,and their attributes including the applicable conditions,advantages and disadvantages are analyzed.Depending on different forms of bathymetric sensor and observation quantity,the advantages and disadvantages of single-beam and multi-beam observation models based on observation range and projected water depth are compared.Finally,several typical terrain matching navigation methods are summarized from the perspective of estimation,and their corresponding model assumptions are given.Considering the actual reference navigation system and underwater environment characteristics,the shortcomings of robustness and real-time performance of existing models are analyzed.2.Studying underwater prior digital mapping method based on multi-beam bathymetric data.Firstly,the bathymetric data processing and mapping process based on multi-beam sonar are given after introducing the structure and working principle of multi-beam sounding system.Three typical acoustic profile models in acoustic tracing are compared and analyzed.Through the actual multi-beam sounding data processing,it is verified that the acoustic tracing based on constant gradient and constant sound velocity profile model can reflect the acoustic propagation process more reasonably than the average sound velocity model,and improve effectively the accuracy of multi-beam sounding.Secondly,aiming at the problem of automatic detection of gross error in multi-beam sounding data,an automatic detection method of outliers in multi-beam sounding based on robust Gaussian process regression was proposed combining the learning ability of Gaussian process regression with the robustness of truncated least square estimation.Simulation experiments and actual multi-beam sounding data processing experiments are carried out,and the results show that the proposed method has good detection ability for large,small and population outliers in the sounding data.Then,aiming at the shortcomings of the existing terrain meshing methods of average weighting,inverse distance weighting and Gaussian weighting,a multi-beam sounding data meshing method based on generalized maximum likelihood estimation was proposed.The results of simulation experiment and field test show the false terrain introduced by outliers can be effectively suppressed,and the real underwater terrain can be characterized more reasonably.Finally,based on the proposed method,a large-area high resolution underwater prior topographic map is constructed.3.Studying a robust and fast terrain matching localization method based on multi-beam bathymetric sonar.Firstly,five key steps of terrain aided navigation are introduced.Based on estimation theory,Cramero lower bound for terrain matching positioning are derived,and a selection scheme of combination measurement number based on Cramero lower bound is given.Then,four terrain interpolation reconstruction methods including nearest neighbor,bi-linear,cubic spline and Kriging interpolations are compared.Experimental results show that there is little difference in the accuracy of the four methods for high resolution map interpolation,but for low resolution map,Kriging interpolation and cubic spline interpolation are better.However,Kriging interpolation requires a large amount of calculation,which is not conducive to the real-time implementation of terrain matching positioning.Finally,from the perspective of robustness and real-time,a robust and fast terrain matching localization method was proposed to solve two problems,one is false localization which is caused by using Gaussian likelihood function as similarity measure,the other is a computational burden of localization which is caused by using grid traversal search method in traditional terrain matching localization method.In the proposed method,Huber function in generalized maximum likelihood estimation is used as the similarity measure function,and differential evolution search method is adopted for fast search because its global quick optimization characteristics.Through computer simulation and off-line field test,the results show that the proposed method can not only effectively suppress the influence of outliers on matching location results,but also ensure the global location optimization with less computation,which is beneficial to real-time localization implementation.4.Studying robust terrain tracking navigation method based on recursive Bayesian estimation.Firstly,the performance of underwater terrain tracking navigation methods based on extended Kalman filter,unscented Kalman filter,point mass filter and particle filter are compared.The simulation results show that four methods have better navigation performance using multi-beam sounding model than single beam sounding model.Moreover,point mass filter and particle filter have higher navigation accuracy and more stable convergence performance than extended Kalman filter and unscented Kalman filter,but particle filter has less computation than point mass filter.Then,aiming at the selection of particle number in standard particle filter,a terrain tracking navigation method based on adaptive sampling particle filter was proposed.The simulation results show that the adaptive particle filter can guarantee the navigation precision and improve the real-time performance of navigation.Finally,a terrain navigation method based on robust particle filter is proposed to reduce the divergence of standard particle filter with measurement outliers.The simulation and offline field test results show the proposed method can effectively suppress the filtering divergence caused by the outliers and improve the robustness of terrain tracking navigation.