IoV-Based Collision Avoidance Architecture Using Machine Learning Prediction

It is not practical to assume that all vehicles must have the communication ability for the purpose of realizing the IoV-based collision avoidance architecture. Hence, we need some extra designs to make up for the deficiency. In the literature, some researches use a camera mounted on the infrastructure at an intersection to implement the collision detection. Particularly, they utilize the real-time object detection and dynamic prediction to predict the future positions of vehicles for the collision avoidance. In this paper, we propose a new method to predict the future positions of vehicles as well, which is based on a well-known real-time object detection project, YOLOv4. By incorporating the concept of vehicle dynamics with our machine learning architecture, our design can estimate the further future vehicular position more accurately and stably. Lastly, the experimental results show the performance of our algorithm for predicting the future vehicular positions and realizing the collision avoidance architecture.