Driver's Intention Identification and Risk Evaluation at Intersections in the Internet of Vehicles

In recent years, the rapid improvement of sensor and wireless communication technologies powerfully impels the development of advanced cooperative driving systems, generating the demands to form the Internet of Vehicles (IoV). With the assistance of cooperative communication among vehicles, the road safety can be greatly enhanced in the IoV. In this paper, we propose a cooperative driving scheme for vehicles at intersections in the IoV. First, the driver's intention is modeled by the BP neural network trained with driving dataset. Then, the identified intention is used as the control matrix of the Kalman filter model, by which the vehicle trajectory can be predicted. Finally, by collecting the information of vehicles' trajectories at the intersections, we develop a collision probability evaluation model to reflect the conflict level among vehicles at intersections. Through obtained collision probability, the driver or the autonomous control unit can determine the next step to avoid the possible collisions. Numerical results show that our proposed scheme has high accuracy in terms of driver's intention identification, trajectory prediction and collision probability evaluation.