Dynamic obstacle avoidance control of autonomous vehicles using LIDAR visual inertial fusion SLAM

In order to improve the success rate of obstacle avoidance for autonomous vehicles, a dynamic obstacle avoidance control method for autonomous vehicles using LIDAR visual inertial fusion SLAM is proposed. Visual Inertial SLAM uses FAST feature extraction and KLT optical flow tracking for feature point extraction and tracking, and utilizes IMU data pre integration to optimize inter frame state variables and restore camera scale information. Correct point cloud deformation through VIO pose data, and align point clouds through feature extraction and matching to solve pose change problems. Based on LiDAR data, environmental information is constructed to determine passable sectors, select the direction of travel, and calculate the final direction of movement to achieve dynamic obstacle avoidance control for autonomous driving. The experimental results show that the obstacle avoidance method proposed in this paper maintains a success rate of over 98% and the longest obstacle avoidance time does not exceed 2s.