Research On LiDAR Map-aided Multi-sensor Fusion For Precise Navigation In Urban Environment

Precise positioning plays a vital role in vehicle-related applications,such as autonomous driving and delivery.Those intelligent driving applications put forward higher requirements for the accuracy,robustness and stability of vehicle positioning.For exiting technologies in positioning,Global Navigation Satellite System(GNSS)can provide all-weather real-time global precise positioning service in open-sky environment;Inertial Navigation System(INS)is an autonomous system with strong anti-interference capabilities,which can realize high-frequency,real-time state estimation;Light Detection And Ranging(Li DAR)has high ranging accuracy and can provide support for Simultaneous Localization and Mapping(SLAM)applications;a priori map can provide accurate environmental information,benefiting the navigation system by adding reliable constraints to the positioning module.To take full advantage of those technologies,we studied the map-aided multi-sensor integrated system in this paper.The main contents and contributions are as follows:(1)Aiming at the discontinuous and unreliable positioning caused by GNSS signal occlusion in urban environments,we proposed a tightly coupled GNSS PPP/INS/Li DAR integration method(GIL)based on multi-state constrained Kalman Filter(KF).This method tightly integrates the raw measurements from multi-GNSS PPP,MEMS-IMU,and Li DAR to achieve high-accuracy and reliable navigation in urban environments.Several experiments were conducted to evaluate this method.The results indicate that the root-mean-square errors(RMSEs)of the position and yaw angle estimation results of the GIL method are improved by 60%compared to the multi-GNSS PPP/INS tightly coupled solution in GNSS challenging environment.Besides,the accuracy of velocity and attitude estimation can also be enhanced with the GIL method.(2)Aiming at the problem that the traditional point cloud registration methods take up a lot of storage space and require a large amount of computation.This paper proposed a 2D point cloud registration method based on geometric features.In contrast to the conventional methods that use the entire point cloud for registration,the propoesd method firstly performs the geometric feature extraction and matching,then 2D pose estimation is implemented based on accurate abnormal matching detection and removal.Experiments show that the proposed method is 20 times more efficient than traditional point cloud registragion methods such as Iterative Closest Point(ICP)and Normal Distributions Transform(NDT).(3)In order to cope with the problem of high loading cost of dense point cloud map,this thesis studies the Li DAR prior map construction method.Based on the pose provided by the POS system,the Li DAR point cloud map and the Li DAR geometric feature map were constructed in this paper.The results indicate that,the average size of the geometric feature map is 3.5 M/km~2,which is significantly lower than 1792 M/km~2 of the point cloud map.The feature map can greatly reduce the computational burden of map loading,meanwhile its production dose not require a huge amount of manpower like high-definition map(HD map),which has advantages in cost and practicality.(4)In order to further improve the stability,reliability and continuity of the multi-source fusion positioning system in urban environments,this thesis incorporates the prior map and the fusion positioning algorithm,and implements the prior map-aided positioning algorithm.The experimental results demonstrate that the error accumulation problem of Li DAR/INS fusion system is effectively mitigated with map constraints,and the map-aided algorithm can maintain decimeter-level positioning accuracy in GNSS denied environments.Besides,in the campus environment with GPS signal occlusion,the positionging RMSE of geometric feature map-aided algorithm is 0.5 m,which have the improvement of 76%in comparsion with the GPS/INS tightly coupled solution.