Research On SLAM Algorithm Of Mobile Robot Based On 3D Lidar

3D laser SLAM technology is one of the important research of outdoor mobile robots.The three-dimensional laser SLAM can estimate the pose of the mobile robot while construct a 3D environment map,which is an important basis for the mobile robot to realize positioning and navigation.Aiming at the cumulative error of laser SLAM,the research is mainly carried out from the method of graph optimization constraint and the method of multi-sensor fusion.The main content of the paper is as follows:(1)The laser point cloud matching algorithm determines the accuracy of the 3D laser SLAM.Through experiments and principles,the iterative closest point matching,normal distribution transformation matching and feature matching algorithms are compared and analyzed.The experimental results show that the feature matching calculation is small and the accuracy is high.So choose feature matching as the basis for research.Aiming at the problem of point cloud matching error increasing with time,it is proposed to construct a pose map model.Set the lidar pose as the vertex,and the point cloud matching result as the order constraint,and performing loop detection based on the Euclidean distance of the vertex to add the loop constraint.Global optimization is performed to reduce the cumulative error.The experimental results show that the method of constructing the pose image can reduce the cumulative error,form a closed loop,and improve the map accuracy.(2)In view of the error in the height direction of the laser matching algorithm,it is proposed to use the RANSAC algorithm to perform real-time ground extraction on the point cloud data based on the globally consistent ground hypothesis.Construct an error equation based on the ground parameters of the point cloud data at different times.Ground constraints are added between them,and a low-frequency ground detection scheme is adopted to reduce the consumption of computing resources.Experimental verification shows that this method effectively reduces the height direction error and occupies less computing resources.(3)Aiming at the situation where the accumulated error is too large to detect loopback,a method of fusion of lidar and RTK is proposed.First,the sensor's coordinate system is calibrated based on the RTK and lidar posture data.Then the RTK data is transformed according to the calibration result to construct an error equation,and RTK constraints are added between the vertices of the pose image.Experimental result shows that this method can effectively solve the problem of unable to loop back.In view of the dependence of the matching algorithm on the initial value,it is proposed to integrate the acceleration of the IMU as the initial value of the match and weight the fusion of the posture of the IMU and the point cloud matching to improve the precision of the posture.Experimental result shows that the fusion of the IMU can improve the accuracy of matching.(4)Build an experimental hardware and software platform,and compare and verify the proposed algorithm.The experimental results show that the proposed algorithm can reduce the error and improve the accuracy,and the calculation amount does not increase significantly.The positioning experiment results show that the proposed three-dimensional map constructed by the improved laser SLAM algorithm can be used for positioning and navigation of outdoor mobile robots.