Research On 3D Laser Navigation Of Mobile Robot Based On Segmentation Matching And SLAM

Three dimensional laser simultaneous localization and mapping(Simultaneous Localization and Mapping,SLAM)technology is one of the most advanced simultaneous localization and mapping technologies for mobile robots.Aiming at the three-dimensional laser SLAM algorithm based on segmentation and matching: the drift generated by the SegMatch class algorithm has accumulated into significant drift over time,this thesis applies the low drift and strong real-time characteristics of lidar odometry and mapping algorithm(lidar odometry and mapping algorithm,LOAM)and its improved algorithm: a lightweight and ground optimized lidar odometry and mapping(Lightweight and Ground-Optimized Lidar Odometry and Mapping,LeGO_LOAM)and advanced implementation of LOAM(Advanced implementation of LOAM,A_LOAM),a combined SLAM method combining the above two kinds of algorithms is proposed.In this thesis,the three-dimensional laser SLAM algorithm and path planning of mobile robot are studied.The main contents are as follows:(1)The method of simultaneous localization and mapping is studied.Firstly,the LOAM class algorithm are described in detail;secondly,the accuracy of map construction,algorithm execution efficiency and loss,map construction time and relative pose estimation of the three algorithms are evaluated;thirdly,the SegMatch class algorithm based on segmentation matching are described in detail;finally,the accuracy of map construction,algorithm execution efficiency and loss,map construction time and relative pose estimation of the two algorithms are evaluated.(2)The combination SLAM scheme of Seg Match class algorithm and LOAM class algorithm is established.Firstly,a new node is written in SegMatch class algorithm to realize the function of tf published by the laser mapping node of the SegMatch class algorithm subscribing to the LOAM class algorithm;then,three kinds of nodes are started simultaneously: the nodes of SegMatch class algorithm and LOAM class algorithm and the new node;finally,the relative pose estimation error of the SegMatch class algorithm based on the scheme is evaluated.The experimental results show that this scheme can improve the relative pose estimation accuracy of Seg Match class algorithm,and is higher than that of LOAM class algorithm.The improved SegMatch class algorithm can achieve closed-loopmapping.(3)The simulation experiment of path planning method based on A* algorithm and dynamic window algorithm is realized.Firstly,the map of C5 plant area and parking lot in the third phase of SIASUN intelligent industrial park is constructed by using optimization algorithm which is SegMatch combined with A_LOAM;then the constructed three-dimensional dense point cloud map is transformed into two-dimensional grid map and published by using octomap server algorithm;then the two-dimensional grid map is subscribed by using move_base navigation function package,and the parameters are changed by using A* algorithm and dynamic window calculation method to implement the overall and local path planning.The experimental results show that the mobile robot model can start from any point on the large-scale outdoor scene map which has been constructed,and can accurately avoid obstacles to reach the set target point.