A new longitudinal car-following control scheme of AVs towards the non-connected situation

Abstract

Connected cooperative driving is known as the promising way to mitigate traffic congestion, enhance driving safety and improve fuel economy. However, before vehicle-to-vehicle (V2V) communication technology became widely applied, vehicles could not always communicate with the front cars due to the uncertainty of vehicle type and communication function. Towards the non-connected situation and making the most of the on-board sensors of the automated vehicles (AVs), an auto-regression (AR) model was adopted to predict the velocity of the preceding vehicle at the next moment, then a new longitudinal car-following control scheme is given from the perspective of cyber physical system to improve the longitudinal following performance. The sufficient condition ensuring a better performance is acquired by local stability analysis and the impact of velocity prediction errors of the AR model is analyzed through a nonlinear partial differential equation. The experiments based on the US-101 dataset and numerical simulations were carried out and the results are in great agreement with the theoretical analysis, which reveals that applying AR model to predicting the velocity of the preceding vehicle at the next moment can improve the car-following performance of AVs without the support of communication devices.