Autonomous car-following approach based on real-time video frames processing

Car-following theory has received considerable attention from the transportation fields in the last decade. Autonomous vehicles are designed to provide convenient and safe driving by avoiding accidents caused by driver errors. However, it is always important to enhance the recognition of driver's driving-style in our roads. With car following models, automated vehicles can imitate the human behavior driving and assure a high safety level on road. All the built-in technologies must be integrated and complemented to achieve these goals. Automated object detection and the following process are one of the main research tasks that must be undertaken for this purpose. In this paper, we design a car following framework for autonomous vehicles using the reinforcement learning technique. The objective is to follow one leader car based on image-frames. To this end, we propose to employ the YOLOv3 object detector to detect vehicles the Q-learning technique to train the follower vehicle to autonomously navigate and follow the leader. Simulation results show the convergence of the model and investigate the behavior of the image-based autonomous vehicle in following its leader.