Through the coordinated control of the main and pre-signal, the contraflow left-turn lane intersection dynamically transforms the opposing exit lane into the entrance left-turn lane. This transformation has improved the capacity of leftturning vehicles. However, during vehicle operation at the intersection, traffic conflicts can easily occur for vehicles making left turns. This safety issue should not be underestimated. Data collection in this study was conducted using drones and camera equipment. This study focuses on left-turning vehicles with different entrance lanes appearing at the intersection at the same time. Characteristic variables affecting the safe operation of vehicles were extracted. The traffic conflict indicators PET, forward conflicts, and rear-end conflicts are used as conflict evaluation criteria. Logistic regression was used to analyze significance and to exclude irrelevant factors. With the utilization of random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and multilayer perceptron (MLP) methods, an optimal binary classification model was trained and the relationship between traffic characteristic variables and traffic conflict is established. The results show that the random forest model can effectively predict the occurrence of traffic conflicts based on the characteristic variables collected by the 10-second time unit, and the overall accuracy is 81.8%.
Research on the driving safety prediction method of left-turn motor vehicles at contraflow left-turn lane intersections
Fourth International Conference on Applied Mathematics, Modelling, and Intelligent Computing (CAMMIC 2024) ; 2024 ; Kaifeng, China
Proc. SPIE ; 13219
2024-07-21
Conference paper
Electronic Resource
English
A Comprehensive Double Index Left-Turn Conflict Model at Contraflow Left-Turn Lane Intersections
DOAJ | 2024
|Signal timing and geometric design at contraflow left-turn lane intersections
DOAJ | 2022
|Wiley | 2019
|