Friction or skid resistance is crucial for road safety as it governs the traction force between vehicle tires and pavement. This study evaluates the performance of various Machine Learning (ML) regression algorithms in predicting traction force and factors affecting the traction force during Locked Wheel Skid Trailer (LWST) tests on New Mexico Interstate Highways. Up to 18 ML algorithms, including Extra Tree Regressor (ET), Gradient Boosting Regressor, and Random Forest Regressor, using 10-fold stratified cross-validation for evaluation have been modeled. The ET model emerged with the lowest RMSE and comparable accuracy to traditional physics-based models. Key findings include that “Load” was the most significant feature with a variable importance score of 0.67. The ET model achieved an R2 value of 0.968 on test data, reflecting strong predictive accuracy. This study demonstrates the effectiveness of ML models in predicting traction force and highlights the importance of specific variables, such as load, in enhancing road safety through precise traction force predictions.
Machine Learning-Based Evaluation of Traction Force in Locked Wheel Skid Trailer on New Mexico Interstate Highways
International Conference on Transportation and Development 2025 ; 2025 ; Glendale, Arizona
05.06.2025
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Speed Adjustment Factors for Locked-Wheel Skid Trailer Measurements
Online Contents | 2010
|Speed Adjustment Factors for Locked-Wheel Skid Trailer Measurements
Transportation Research Record | 2010
|Transportation Research Record | 2025
|Comparative Evaluation of Locked-Wheel and Fixed-Slip Skid Systems
Online Contents | 2013
|