AUV Bathymetric Simultaneous Localization And Mapping

Bathymetric simultaneous localization and mapping(BSLAM)technology could provide location results for autonomous underwater vehicle(AUV)without accumulative error.According to the state estimated and topography observed,AUV could correct its estimation state detect loop closures via matching process between historical measurement and new observation.At this time,AUV would build the new map or modify the historical map,in order to construct the accurate seabed topography map and use it for location and mapping simultaneously.BSLAM technology can measure and model the environment while providing accurate navigation results for AUV,and it is an important technical means to realize the AUV long-term and autonomous underwater navigation.The main work of this paper includes:This paper studied the pose graph construction in the BSLAM.Starting with the raw data of multi beam sonar,this paper analyzed the spatial location calculation and filtering of multi beam data,and proposed to represent the size of complex terrain area contained in the map by the difference of Normals(DoN),and this was an important basis for submap division.The multi beam sonar could only measure the seabed topographic swath in linear sequence,and this meant no information redundancy between adjacent frames.So,this made it difficult to estimate the AUV motion between adjacent frames.This paper proposed the weak data association construction method to solve this problem.In this method,a sparse pseudo-input Gaussian processes(SPGPs)model was built to estimate the seabed topography,and then the AUV motion between adjacent frames could be estimated by comparing the estimated terrain with the measurement.To improve the computational efficiency,the searching area size of loop closure detection was delimited,and the loop closure detection process was realized by template matching method.In the BSLAM pose graph optimization process,to deal with the low belief of weak data association in the pose graph,the moments associated with loop closures were extracted as the key frames,and the pose graph optimization algorithm combined global and local trajectory corrections were proposed to modify AUV states on key and non-key frames.In the global trajectory correction,the continuity equation of navigation deviation on key frames was proposed,and the correction problem of navigation deviations on key frames was transformed into the least squares problem and then solved.The local path correction answered how to assign the navigation deviations on the key frames to the whole path.This paper constructed the spring model of navigation deviations and proposes the terrain correlation method to deal with this model.On this basis,the direction of navigation deviation was considered and an improved terrain correlation method was proposedIn regard to the robustness extension of BSLAM,this paper solved the problem of the identification and rejection of invalid loop closures.In pose graph construction,this paper proposed an invalid loop closure idenfication method considering the residual distribution,and this method identify the validity of every single closure by the distribution of residuals in the spatial and frequency domains.In pose graph optimization,this paper put forward both a voting algorithm and a multi-window consistency method.The invalid loop closures were identified sequentially by the coupling relationship between two or all closures.In this paper,a robust BSLAM system was designed and both numerical simulation and online sea trail experiments were carried out.In the numerical simulation experiment,the visual simulation system was built,and the multi beam data are simulated via intersection line detection.The pose graph construction,optimization and the robust BSLAM method were verified in the numerical simulation,and the performances of the voting algorithm and multi-window consistency method were compared.In the online sea trail experiment,the performance of the robust BSLAM system was tested in the both still and rough sea conditions.To deal with unstable BSLAM results under rough sea conditions,the multi-window consistency method with filter was proposed.This paper also analyzed the feasibility of registration error distribution estimation according to the distribution of terrain standard deviationAn autonomous map extension method for AUV was proposed.In the autonomous map extension method,AUV located itself by terrain matching process in known partial prior map,and built the map according to the location results and BSLAM result.This method mainly includes two aspects:path planning in a known environment and positioning in an unknown environment via robust BSLAM algorithm.The latter has been fully studied in the previous chapters of this paper,and its performance have also been proved.Started with the map extension mission decomposition and the environment modeling,this paper deduced the Markov distance calculation method considering the terrain complexity,and proposed to improve the new node generation method by using area of interested and roulette method.In this paper,A*and Rapid-exploration Random Tree*(RRT*)algorithm for terrain aided navigation were proposed,and the performances of both algorithms were analyzed through simulation experimental results.In order to verify the performance of AUV autonomous map extension method,the simulation experiment was carried out.