| As research on unmanned ground vehicles is moving towards new areas of civilian and diversified development,Simultaneous Localization and Mapping(SLAM)as the core technology to realize unmanned vehicle autonomous positioning,is developing to wider range,higher precision,lower cost and more complex environment.In this paper,we study the SLAM technology based on single LIDAR for park patrol,including the construction of pose graph(front-end)and graph optimization algorithm(back-end).The following contributions and innovations are obtained:1.A multiple frame point cloud matching(Multiple-CSM)algorithm based on prior information of the environment is proposed.This method is based on multi resolution correlative scan matching algorithm,using variance estimation for the environment prior information,building the grid query table with multi frame point cloud.Multiple-CSM effectively improves the matching accuracy and the ability to adapt to complicated environment.2.A loop-closure detection method based on Monte Carlo combined with normal distribution transform(MC-NDT-Matching)is proposed.MC-NDT-Matching takes the advantage of the accuracy of Monte Carlo localization and the fast convergence of NDT in large scale point cloud registration.The two methods are combined together to improve the accuracy of loop-closure detection through the selection of candidate points and to construct a reliable pose graph for the back-end optimization algorithm.3.An improved iSAM(Incremental Smoothing and Mapping)algorithm based on error detection is proposed.The loop-closure rejection and iSAM are combined to determine whether the newly added constraint should be rejected or not by detecting the sequence error produced by the optimization.In addition,the improved iSAM algorithm takes the advantage of covariance estimation,and prior information is introduced into optimization to improve the accuracy of the global optimization.Based on the front-end matching,loop-closure detection and back-end optimization,a large range,high precision,low cost Simultaneous Localization and Mapping based on single LIDAR is realized. |
