Research On SLAM Method Based On Tight Coupling Of IMU And Lidar

Localization and Mapping(SLAM)is an important method to realize environment awareness,real-time localization and path planning for intelligent mobile robots,and is also a key technology for intelligent mobile robots.Aiming at the problems such as large shadow and low positioning accuracy of maps built by intelligent mobile robots in complex scenes,this work proposes a SLAM algorithm integration and optimization method based on two sensors,lidar and inertial measurement unit(IMU),to meet the requirements of accurate positioning and mapping in complex scenes.The main research contents are as follows:(1)Due to the motion distortion of lidar and the gravity vector drift of traditional lego-loam algorithm,the lego-loam algorithm(imu-lego-loam algorithm)which is tightly coupled with lidar is adopted.Based on the data fusion process of IMU and lidar,the motion state is estimated by IMU.To eliminate the motion distortion of lidar point cloud;Meanwhile,IMU and LIDAR were used to construct a joint optimization function to constrain the gravity direction of position attitude estimation,so as to solve the gravity vector drift problem caused by sparse ground data.(2)In complex scenes,although the IMU-LEGO-LOAM algorithm has been greatly improved in operation efficiency,its positioning accuracy is still relatively low.Therefore,a tightly coupled IMU and lidar hybrid scanning matching algorithm(IMU-ICP/LEGO-LOAM algorithm)is further proposed,including: Lego-Loam algorithm is adopted for the point cloud matching pattern in the scene with rich structural information,while ICP algorithm is adopted for the point cloud matching pattern in the scene with less structured information.In addition,the position and pose estimated by IMU were used as the iterative initial solution of point cloud matching to accelerate the system running speed.At the same time,the system error equation of IMU and LIDAR is constructed to realize the joint iterative optimization,so as to improve the positioning accuracy of the system.(3)for intelligent mobile robot in the case of mileage increasing,the cumulative error increases,positioning accuracy decline and increase map ghosting,by using the current position and the historical position of loopback detection,and add the loopback constraint to remedy the registration form,effectively restrain the increase of the cumulative error,and improve the intelligent mobile robot positioning accuracy.At the same time,the front-end odometer relative pose,loop detection relative pose and IMU estimated relative pose were fused to achieve back-end optimization,which effectively reduced the map ghosting and further improved the positioning accuracy of the system.In addition,according to the field test results: The relative pose errors of the traditional LEGO-LOAM algorithm,IMU-LEGO-LOAM algorithm and IMU-ICP/LEGO-LOAM algorithm are 0.4553 m,0.3422 m and 0.3049 m,respectively,and the cumulative errors are 18.8048 m,3.4467 m and 1.9219 m,respectively.The running time is97.6ms,60.2ms and 86.1ms,respectively.Based on this,we can know that IMU-ICP/LEGO-LOAM algorithm,compared with traditional LEGO-LOAM algorithm and IMU-LEGO-LOAM algorithm,has the best mapping effect and positioning accuracy under the premise of ensuring the running efficiency of the algorithm.