Pose Estimation Of Mobile Robots Based On Fusion Information Of Lidar And Stereo Visual Inertial

With the popularity of mobile robots in different fields,the pose estimation of robots has became a research hotspot.In order to achieve accurate pose estimation of robots in different complex environments with computation efficiency and robustness,this paper made three improvements based on the fusion of Lidar,stereo,and IMU in the framework of extended Kalman filter(EKF).First,this thesis improves the conventional stereo visual inertial odometry(VIO)with two stage EKF.The first stage of this EKF-based algorithm performs the fusion of accelerometer and gyroscope while the second performs the fusion of stereo camera and IMU.Compared to previous stereo VIO based on EKF,this method employs the information of accelerometer to correct the gyroscope error.This improvement is validated through the public dataset Eu Roc.The simulation results show that the accuracy of the proposed VIO based on the two-stage EKF achieves 15.7% higher than that before this improvement.Second,this thesis improves the application of RANSAC in the screening of visual matching points.By using the rotation compensation of the IMU,the image of the previous frame is rotated to the current frame.The number of cycles of RANSAC algorithm can be reduced by employing the distance ratios of the different feature points on the normalized planes of the two image frames are approximately equal.This improvement is validated through the public dataset Eu Roc.The results show that the time consumption of the proposed RANSAC achieves 74% lower than that before this improvement.Finally,this thesis employs the information of the lidar to enhance the robustness of the stereo VIO in texture-less and dark environment.By extracting corner points and surface points from the laser point cloud,and finding matching points in adjacent frames according to the point line and point plane ICP,the observation constraint formed by matching point pairs is added to the VIO as a measurement update.Through experiments,it is verified that the fusion of the lidar improves the robustness of the stereo VIO in complex scenes.