Spare Pose Graph Decomposing And Optimizing SLAM Algorithm Based On Composite Clustering NDT

In this thesis,we focus on the mapping speed and precision in SLAM(Simultaneous Location and Mapping),and present a spare pose graph decomposing and optimizing SLAM algorithm based on composite clustering NDT(Normal Distribution Transform).Composite clustering NDT method is used in the period of data fusion,which composes density-based clustering and K-means clustering to aggregate the points with similar local distributing feature.It takes SVD to judge the suitable degree of one cluster for further division.Meanwhile,to avoid the radiating phenomenon of LIDAR in measuring point distance,we propose a method based on trigonometric to measure the internal distance.The clustering method could ensure the expression of LIDAR's local distribution and matching accuracy.With the scale increasing of pose graph,optimizing a large-scale pose graph is the bottleneck of graph-based SLAM.In this thesis,we propose an optimizing method basing on the spare decomposition of pose graph.With the extraction of the Single-chain and the Parallel-chain,the pose graph is decomposed into many small sub-graphs.Compared with directly processing the original graph,the speed of calculation is accelerated by separately optimizing the sub-graph,as the computational complexity is increasing exponentially with the increase of graph scale.Finally,our algorithm is validated with the public dataset of mobile robot.The result demonstrates that compared with traditional algorithms,our method is more accurate and more stable.