| As the society’s demand for autonomous navigation products increases,the research on autonomous driving technology has continued to deepen.Among them,positioning technology is the prerequisite for the realization of autonomous driving technology.At present,there are many solutions to the positioning problem.Such as track calculating positioning algorithm based on IMU inertial navigation or wheel odometer can achieve short-term positioning,however it will accumulate errors over time;environment-aware positioning algorithms based on Li DAR or vision can accurately identify the current pose state,however it will lose the global pose direction over time.Therefore,based on the analysis of the shortcomings of the single-sensor positioning algorithm,this paper adopts the error state Kalman filter algorithm(ESKF)to fuse the Li DAR and inertial navigation unit(IMU)information to obtain the pose estimation state.At the same time,an improved ICP point cloud frame matching algorithm is proposed.Through the voxelization of the point cloud data grid and down-sampling processing,the interference of redundant point clouds is reduced,and the nearest point pair of the laser point cloud is secondarily selected to improve the speed and accuracy of point cloud registration between frames.Finally,a 3D point cloud map is constructed through ESKF fusion positioning information.The first step of optimized iterative matching is achieved through inter-frame matching.This method outputs higher frequency laser odometer information,and then adds ESKF algorithm to it to fuse the laser odometer and the optimization factor after IMU pre-integration.Then the system outputs the odometer frame conversion information at a higher frequency,and finally uses the current point cloud frame to local point cloud matching method to establish a point cloud map.Through experimental comparison with the traditional LOAM mapping algorithm,the results show that the accuracy and mapping effect of the ESKF algorithm combined with the IMU positioning and laser odometer are higher than that of the LOAM algorithm.It can be concluded that the positioning accuracy of ESKF algorithm which fuses multi-sensor information can make up for the lack of state estimation of a single sensor state estimation algorithm. |
