Queue length estimation model for mixed traffic flow of intelligent connected vehicles and human-driven vehicles

A dynamic queue length estimation model based on probability statistics and Bayesian theorem was proposed, to solve the problem of queue length estimation at intersections with mixed traffic of intelligent connected vehicles (ICVs) and human-driven vehicles (HDVs). Firstly, taking into account factors such as the position, speed, and penetration rate of ICVs in the queue, models for estimating the queue lengths of observable and unobservable queues, as well as the penetration rate, were constructed. Real-time estimation of queue lengths and penetration rate was achieved through iteration. Then, the distribution characteristics of ICVs in the queue under different penetration rate conditions were simulated using random seeds. The estimation accuracy of the model under different traffic conditions was analyzed. Comparison analysis with existing models showed that, under low penetration rate conditions of ICVs (10%) during off-peak hours, the average absolute percentage error (MAPE) of the proposed model was 29.35%, while the existing model had an MAPE of 59.68%; during peak hours, the MAPE of this model was 26.50%, compared to 34.66% for the existing model. Under high penetration rate conditions of ICVs (90%) during off-peak hours, the MAPE of this model was 6.90%, while the existing model had an MAPE of 17.85%; during peak hours, the MAPE of this model was 1.45%, compared to 1.05% for the existing model, with similar errors. The proposed queue estimation model for mixed traffic of ICVs and human-driven vehicles has better estimation accuracy under both low and high penetration rate conditions. © 2024 Zhejiang University. All rights reserved.