Measuring pavement side force coefficient using vehicle on-board diagnostic data

This is the authors' abstract. We don't add key points for this paper.

Road traffic accidents are a major global concern, with many caused by reduced pavement skid resistance due to surface degradation. Traditional assessment methods face challenges in balancing cost, efficiency, and coverage. This study proposes a novel approach using On-Board Diagnostic (OBD) data from a fleet of vehicles as a distributed network of mobile sensors to evaluate skid resistance. A set of parameters is developed and validated using CarSim simulations, with certain parameters estimated via recursive least squares (RLS) and Unscented Kalman Filter (UKF). Machine learning algorithms, including a PSO-BP neural network model, are employed to assess pavement skid resistance. Validation on Shanghai Puwei Highway shows the model's high accuracy, with an RMSE of 1.7946 and an R² value of 0.86. The model enables real-time, large-scale monitoring of road safety, offering a cost-effective and efficient alternative to traditional methods.