Cyber-Physical CNN Prediction-Assisted Feedback VSL Control for Mixed Traffic

This study develops a cyber-physical convolutional neural network (CP-CNN) prediction method inspired by image three-primary colors. By incorporating CP-CNN, a feedback variable speed limit (VSL) control is established to enhance traffic efficiency in the vicinity of traffic signals under mixed traffic environment. In particular, the CP-CNN model can predict traffic volume by extracting the vehicle’s physical features to establish cyber feature images of the vehicle group. Furthermore, the feedback VSL integrated with the CP-CNN can better adjust upstream mixed traffic flows. The comparative experimental results verify that the proposed CP-CNN model outperforms that under the existing DL-based predictive model in terms of prediction accuracy. Additionally, under the dense flow and a connected-and-automated (CAV) penetration rate of 10%–50%, the CP-CNN-based VSL control can improve the traditional transportation systems, reducing the average waiting time by 19.1% and the average travel time by 10.4%.