Research On Unmanned Vehicle Location And Navigation Algorithms For Autonomous Parking

With the continuous growth of car ownership,autonomous driving relies on the rapid development of technology,the Internet,big data,etc.,in order to solve the problems of time consumption and prominent contradictions in the parking process,autonomous valet parking has gradually become driverless Research and application hotspots.Unmanned vehicle positioning and navigation mapping based on autonomous parking scenarios is a crucial step in autonomous valet parking.In this paper,a high-precision 3D laser point cloud based on lidar scanning is combined with Inertial Measurement Unit(IMU)data and Global Navagation Satellite System(GNSS)system data to map the autonomous parking scene.The error state Kalman filter method is used to fuse multi-source sensors to solve the current low accuracy of odometer in parking scenes of pure lidar odometers,and at the same time use inertial navigation information to correct long-time GPS in autonomous parking scenes The positioning deviation and the movement distortion of the carrier caused by the signal loss can improve the accuracy of the pose.According to the strength of the GPS signal,the fusion of different sensor combinations is carried out,so that the algorithm can run relatively stably.Secondly,based on the point cloud matching algorithm,according to the line,surface and ground features in the autonomous parking scene,the feature point cloud is extracted for registration and ground detection and extraction,reducing the redundancy between key frames The registration between ground points enables the algorithm to meet the real-time performance of the algorithm while correcting the pose of the key frame.Finally,in order to verify the effect of the algorithm in this paper,a smart vehicle test platform was built to simulate the scenes of good GPS signal and poor GPS signal in the autonomous parking scene in multiple parking lots on campus,and collect multiple sets of lidar point cloud data.The algorithm performs field scene verification and compares it with the current scheme of laser synchronous positioning and mapping in the field of unmanned driving.Experiments show that the algorithm can effectively reduce the cumulative error of the mapping process,and at the same time achieve more accurate vehicle self-positioning,and provide good a priori conditions for autonomous parking planning and control algorithms.