Research On SLAM Localization Algorithm Of Mobile Robot Based On Tight Coupling Of Lidar And IMU

With the development of autonomous mobile robotics,scholars in related fields pay much more attention to this field.SLAM is the key technology of mobile robot localization.There are many solutions of mobile robot positioning,such as wheeled odometry and inertial guidance for position projection but the cumulative error of the algorithm becomes larger after a long period of operation.Lidar or vision can position by external environmental information which are more accurate in identifying the position state,but performance is bad in the large range positioning.Therefore,after considering the shortcomings of a single sensor in this thesis,a tightly coupled Lidar and IMU scheme is chosen to estimate the pose from the data between two sensors and to optimize it.Firstly,the raw data of IMU is used preintegration to estimate the pose,and correction of point cloud distortion;secondly,according to the obtained point cloud to extract features,and store them in categories,they are divided into plane points and edge points;thirdly,analyze the problem that the traditional ICP matching algorithm needs to provide initial values and iterative optimization takes a long time in the feature matching stage.Propose an improved scheme to provide initial values by IMU pre-integration,voxelization of point cloud is processed to increase the iterative speed of the algorithm and using point cloud matching module to build a local map;finally,match a frame with the local map to get a distance,the distance and the IMU pre-imtegration create an optimization function.The datasets used in the experiments are the KITTI dataset and the self-recorded park dataset.The results show that the absolute positional error of the improved algorithm is 2.73 m and the average value is 1.14 m when using KITTI dataset,the error values is better than Lego-LOAM and LIO-Mapping,positioning accuracy increased by 34%.The processing time of single frame point cloud of the improved algorithm is282.3ms,which meets the real-time requirements of odometer.