Technical Paper
A high-precision autonomous localization system for autonomous parking scenes based on multi-sensor information fusion
2024-04-09
2024-01-2843
Aiming at the long-term loss of localization in Autonomous Parking Scenario and the cumulative drift problem in SLAM on-line mapping and localization, this paper establishes a GPS、INS、SLAM multi-system fusion localization framework to realize centimeter-level localization with wide scene adaptability under multi-scale unification. In this paper, Lidar and Inertial Measurement Unit (IMU) are loosely coupled to build the point cloud map in the parking lot, and the IMU pre-integration information is used to provide the rough pose of the point cloud frame, and the point cloud distortion is removed to extract the line and surface features for pose estimation. The back-end optimization map is aligned to the GPS coordinate system using GPS and IMU information to initialize vehicle pose, NDT registration of line and plane features to calculate vehicle pose, IMU pre-integration information is used as initial value of registered pose.