An Autonomous Vehicle Approach Based on Improved Target Detection and Trajectory Tracking

Context perception and path planning and tracking control are important parts of autonomous driving technology, which can determine the optimal path for the vehicle and ensure that the vehicle travels along the path accurately. However, at present, autonomous vehicles mainly face two problems: on the one hand, the traditional object detection algorithm is prone to errors in the detection of dynamic targets, and the matching between image feature points is prone to errors, so the detection accuracy is low. On the other hand, due to the constant changes in the external environment, absolute tracking control under a given desired trajectory can not meet the requirements of safe driving of driverless cars. Based on this, this paper proposes a dynamic target detection method combining ORB-SLAM3 algorithm and YOLOv5s algorithm. In this scheme, Hybid A* algorithm is used for path planning and model predictive control (MPC) controller model is used to realize vehicle trajectory tracking, which not only improves the accuracy of target detection, but also ensures the safety of vehicles in the process of trajectory tracking. Through dataset testing and Simulink/CarSim simulation experiments, the good object detection and path tracking capabilities of this scheme are verified.