Research On Application Of Bayesian Inversion In High-accuracy Underwater Localization

As the underwater acoustic localization(UAL)technology keeps developing,UAL systems nowadays have achieved some basic functions of the Global Navigation Satellite System(GNSS),such as localization,navigation and surveying.However,the complicated ocean environment,multi-type system bias and model approximation error decrease accuracy and precision of UAL and make them several orders of magnitude lower than the GNSS.This problem has already drawn back the performance of seafloor geodesy.Basing on the demand of high-precision UAL from seafloor geodesy industry,this thesis develops algorithms to compensate background information errors and model approximation errors with Bayesian inversion theory.It does not only realize localizing a seabed acoustic transponder accurately under an inaccurate sound-speed profile(SSP)in the water column,but also improves the accuracy of a long baseline(LB L)UAL for a maneuvering acoustic source with inaccurate reference coordinate information.An accurate forward model for signal travel-time is significant to inverse for target coordinates.Thus,a high-precision UAL model based on acoustic ray travel-times is established using the ray-tracing theory,which can correct the ray bending problems in the UAL effectively.Further,an analytical expression of the posterior probability density(PPD)for target coordinates is obtained under the linearized Bayesian inversion strategy.A transponder geodetic coordinates and SSP bias joint inversion algorithm is developed to solve the inaccurate transponder localization problem with a biased SSP.This algorithm treats SSP bias as an additional unknown parameter and it is eastimated jointy with transponder coordinates.To decorrelate the transponder depth and SSP bias,observation trajectorys with multiple trade-off angles and precise depth sensor data are used for better data and prior information,which helps the inversion stay robust.Simulation and lake trial results show that this method is effective in compensating SSP bias while localizing a transponder and performs better than the one depends on symmetric observations only.In the lake trial data processing,the localization accuracy of transponder depth is improved to 0.004±0.003 m from 0.046± 0.003m,which is a strong support for this joint inversion method.Furthermore,a transponder geodetic coordinates and SSP temporal variation joint inversion algorithm is developed to address the inaccurate transponder localization problem under temporal varied SSPs.The method is based on dividing observed acoustic travel-time data into time segments and including depth-independent SSP variations for each segment as additional unknown parameters to approximate the SSP temporal variation.SSP variations are estimated jointly with transponder coordinates,rather than calculated separately as in existing two-step inversions.Simulation and sea trial results indicate that the joint inversion impoves the deficiency of the "two-step" method and estimate both transponder location and SSP variations correctly.In the sea trial data processing,the averaged errors for transponder depth and SSP variations inversion are decreased to 0.063±0.032 m and 0.056±0.026 m/s,respctively.Moveover,the computational efficiency is severly improved.To solve the inaccurate reference coordinates problem caused by transponder locations errors in a LBL system,an acoustic source trajectory and transponder locations joint inversion algorithm is applied in the thesis.The reference coordinates information is supplied by Bayesion prior for transponder locations with uncertainties,and they are estimated together with source locations.Simulation and sea trial results show that this algorithm estimates the posterior credibility region for estimates correctly.The influence from transponder locations errors is alleviated so that the source trajectory can be much better localized.In the sea trial data processing,the LBL localization error is 2.5±0.9 m after applying this method,which is much lower than the one ignores the errors in transponder locations(6.9±2.8 m).The acoustic source is assumed static in the source and transponder joint inversion algorithm(static-model),which causes localization errors when the source is actually maneuvering.To address this problem,an joint inversion algorithm is developed to compensate for source motion during the interrogation-reception time interval between the source and transponders of the LBL system using only acoustic timing measurements.The method is based on including travel-time corrections as additional unknown parameters in the inference,constrained with prior estimates determined by interpolating the vehicle location at interrogation time instants from the results of an initial localization based on the static-model.Simulation studies and lake trial indicate that the motion-compensated localization method corrects the inaccurate usage of the one-way travel-time in the static-model,therefore it performs much better than the static-model localization method.Data processiong results from the lake trial are also presented,with averaged acoustic localization errors for a surface maneuvering source reduced from 19±14 cm for the static-model localization to 3.4±1.4 cm for the motion-compensated localization,which successfully achieve the goal to compensate the source motion influence in a LBL system.