Abstract:Aiming at the autonomous parking environment in which the map building and localization accuracy are limited by using only LIDAR sensors, a map building and localization method I-LOAM based on the inertial measurement unit (IMU) tightly coupled with LIDAR is proposed for autonomous vehicle parking scenarios.Firstly, IMU pre-integration of the point cloud data is performed at the front-end, point cloud preprocessing to remove the ground point cloud and reduce the point cloud scale to ensure the efficiency of laser odometry. Secondly, an S-ICP algorithm based on sample consensus initial alignment (SAC-IA) coarse processing and optimized iterative closest point (ICP) fine alignment is proposed, which complements the tightly coupled positioning algorithm with IMU and LiDAR to provide the best solution for autonomous parking. The S-ICP algorithm is complementary to the tightly coupled IMU and LiDAR localization algorithm, providing a more flexible and accurate map building and localization solution for the autonomous parking system. Then, the map is constructed by fusing IMU information, laser odometry and loopback detection information at the back-end. Compared with the LeGO-LOAM algorithm, the proposed algorithm’s rms error is reduced by 45%, 3% and 6% in outdoor, indoor and straight road scenarios, respectively, with better accuracy and robustness, which provides an accurate and reliable solution for map building and localization tasks in autonomous parking environments.