Improved Bathymetric Particle Filter SLAM

Simultaneous localization and mapping(SLAM)is an important technology to realize the positioning of autonomous underwater vehicles(AUVs)and construct the surrounding environment without relying on the support from external acoustic arrays and priori map information.This dissertation aims to realize AUV underwater long-time autonomous and accurate navigation,uses a multibeam sonar as the terrain measurement sensor to conduct the study of the particle-filter-based bathymetric SLAM(BSLAM)and to improve the precision and the efficiency of the particle filter BSLAM methods.The major contributions of the dissertation are:1)To solve the imprecision particle weighting problem when the number of input valid bathymetric data is limited and reduce the computational costs of the map update of a grid-mapbased particle filter BSLAM,a particle filter BSLAM method based on trajectory map representation is proposed.A particle weighting method based on submap matching is designed to achieve accurate update of particle weights.2)A particle filter BSLAM method based on mean trajectory map representation is proposed.To solve the real-time optimal trajectory output problem and particle trajectory oscillation problem using large process noise,a mean trajectory map and the corresponding update method are designed.To improve the robustness of the algorithm,a loop closure consistency determination method based on hierarchical clustering is designed to reduce the influence of invalid loop closures on navigation results.3)A particle filter BSLAM method based on graph optimization theory is proposed.To solve the problem that particle filter BSLAM cannot update the historical states of particles,a graph-based particle trajectory update method is proposed to improve the accuracy of the algorithm.To solve the real-time optimal trajectory output problem of particle filter BSLAM,a graph-based optimal trajectory construction method is designed to realize the real-time output of accurate mapping results.The playback experiments using simulated and sea trial datasets were conducted to verify the effectiveness of the proposed methods.The experiments show that the accuracy of a proposed method is 60% higher,and the computational time is 15% lower than an existing BSLAM method.The proposed BSLAM methods can meet the underwater positioning and mapping requirements of AUVs.