Research On Starting Position Of AUV Terrain Reference Navigation

Terrain reference navigation(TRN)provides Autonomous underwater vehicle(AUV)bounded-error position information relative to the environment.The Terrain reference position(TRP)can be calculated by matching the real-time measurement terrain which is obtained from the sounding equipment on the AUV against the Digital Elevate Map(DEM).Based on the existing research results and conclusions,the TRN technology is further improved.The main work is summarized as follows:The observation model of TRP under unknown tidal range and measurement error is proposed,and the problem of confidence interval estimation for TRP is solved.Because terrain surface can't be described by parametric equation and TRP points are independent,it is very difficult to estimate the confidence interval of TRP.In this paper,the square sum function of location residuals is regarded as a quadric surface function,and the TRP point is regarded as the parameter to be estimated.The confidence interval estimation problem of TRP is described as the problem of solving the confidence interval of quadric surface parameters,so that the confidence interval estimation problem of TRP can be solved.Furthermore,the error analysis method of TRP is extended to the study of terrain adaptation,and the optimal segmentation method of adapted and non-adapted areas under the premise of grid-based prior topographic map is proposed.The grid-adapted map will be suitable for solving the optimal path search problem in TRN path planning.The problem of improving the accuracy and stability of TRP is studied.The main purpose of this part is to solve the problem of initial alignment of TRN.The motivation of initial positioning is to solve the problem of slow convergence or divergence of filtering in the initial stage of TRN.In order to improve the stability of TRP and TRN results,this paper proposes to reduce the negative impact of erroneous terrain information on TRP and TRN by screening and eliminating large errors and distorted measurement points in measurement data.Particle initialization for particle filter TRN.the positioning information provided by the reference navigation deviates from the reality seriously when the cumulative error of the initial reference navigation is large.If the initial particle is initialized by the reference navigation information or TRP information,the coverage of the initial particle will be too large,the probability of self-similar terrain will increase,and the particle filter will be difficult to converge or even diverge.Considering that the errors of TRP is bounded and time-independent,a method of fast convergent particle filter is proposed to initialize TRN by using the effective positioning points of TRP.The algorithm constrains the initial particles in a smaller range through the confidence interval of TRP,thus improving the convergence speed of filtering.The problem of initial position estimation of TRN with large initial positioning deviation is studied.A fusion location method of multi-TRP points is proposed to estimate the initial positioning of TRN.Because of the accuracy of TRP can be affected by terrain features and measurement errors and there is almost no correlation between the measured terrain of the TRP,the TRP position will be stochastic jumping,which also makes it difficult to obtain stable positioning results in the initial stage.In this paper,an information fusion method is proposed by considering the recursive relationship between the reference navigation points.And the linear fusion method of multi-TRP points and the non-linear fusion method of multi-TRP points are studied to fuse the positioning information of multiple TRP points.