Localization And Mapping Algorithm For Driverless Car Based On Fusion Of Lidar And IMU

In recent years,the autonomous driving industry has developed rapidly,and related technologies have become increasingly mature and gradually commercialized.As the basic function of autonomous driving,localization and mapping is the basis for autonomous vehicles to complete environmental perception,decision-making and path planning,and plays an important role in autonomous driving technology.The localization and mapping methods based on a single sensor have shortcomings in terms of accuracy and robustness.For example,Li DAR will degrade in scenes lacking of structural features,cameras are more sensitive to illumination changes,and the direct integration of IMU measurement will lead to divergence of results,GNSS will have signal deterioration in some scenarios due to occlusion and other reasons.In this paper,combined with the advantages of Li DAR which has accurate ranging capacity and is insensitivity to illumination changes,and IMU w hich has high measurement frequency and is independent of external environment,this paper designs a localization and mapping algorithm based on fusion of Li DAR and IMU information,which can achieve high-frequency,real-time,robust,accurate ego-motion estimation while building a point cloud map of the surrounding environment.The main research work of this paper is as follows:1.Use the high-frequency,real-time IMU pre-integration technology to add IMU observations into the optimization framework,fuse Li DAR measurements and IMU pre-integration measurements in the IMU pre-integration factor graph,and use the IMU pre-integration measurements to obtain real-time IMU preintegration odometery.2.Using the real-time IMU odometry to de-distort the laser point cloud,segment the ground points in the undistorted point cloud in an efficient way,extract the structural feature points in the undistorted non-ground point cloud using the principal component analysis method,and use the linear multi-metric least squares ICP The method constructs the incremental equat ion and then realizes the solution of the Li DAR pose,and detects and processes the degradation by analyzing the structure of the incremental equation.3.The loop closure detection module uses both RNN search based loop closure detection and Scan Context scene recognition based loop closure detection.This enables the algorithm to correct accumulated drift through loop closure detection even when running in a large range of scenarios.4.Finally,the algorithm is experimentally verified on multiple public data sets,and compared with other classic SLAM algorithms.The experimental results show that the algorithm in this paper can still ensure accuracy and robustness in a wide range of scenarios.