Researchers have developed a novel AI-driven framework using the XGBoost algorithm to accurately evaluate the skid resistance of asphalt pavements under various conditions. Published in Smart Construction, this breakthrough achieves over 90% prediction accuracy, offering a smarter and more adaptive approach to enhancing road safety and maintenance.
Skid resistance is a critical factor for asphalt pavement durability and traffic safety, particularly under wet or extreme weather conditions. However, traditional evaluation methods, such as sand patch measurements or pendulum tests, often struggle with generalization across different pavement types and face challenges in quantifying multi-scale texture features collaboratively. To overcome these limitations, the Sustainable Asphalt Research Group from Universiti Sains Malaysia (USM), in collaboration with institutions from China and Australia, has introduced an intelligent assessment framework for asphalt pavement anti-skidding performance based on the XGBoost machine learning algorithm.
The study focused on three widely used asphalt mixture gradations: AC13, AC20, and SMA13. The field data for these gradations were obtained from the Shi-Wu Expressway in Hubei Province, China. Using a self-developed pavement texture acquisition device, the team captured high-resolution images under strictly controlled lighting and shooting parameters. Simultaneously, the sand patch method was employed to measure the texture depth at corresponding locations, ensuring synchronized acquisition of both image-based texture characteristics and physical anti-skidding performance indicators. These images were processed via MATLAB to extract multi-scale texture features, including statistical characteristics (energy and entropy) and fractal properties (fractal dimension and multifractal spectrum width Δα). A total of 1,800 datasets were constructed, with 70% used for training and 30% for testing.
"The XGBoost model demonstrated exceptional performance, with test set accuracy of R² > 0.9 and RMSE < 0.06," said Yu Zhao, the first author of the study. "This confirms its capability to assess pavement skid resistance accurately and effectively, surpassing conventional methods that often lack adaptability."
Results revealed that the model's performance varies with gradation type: AC13 fine-graded mixtures achieved the highest prediction accuracy, followed by SMA13, while AC20 coarser mixtures showed slightly lower generalization but still met practical requirements. Feature importance analysis identified fractal dimension as the core predictive factor, emphasizing the role of texture complexity in skid resistance.
In validation tests using British Pendulum Number (BPN) measurements, the model maintained an R²> 0.9 demonstrating reliability in field conditions. "This technology has the potential to be integrated into automated pavement inspection systems in the future, enabling real-time skid resistance monitoring without disrupting traffic," explained corresponding author Prof. Mohd Rosli Mohd Hasan. "It will support customized maintenance strategies for different pavement types, potentially extending service life and reducing accident risks.
This framework provides significant theoretical and technical support for intelligent detection and gradation-adaptive evaluation of pavement skid resistance, paving the way for safer roads and more efficient maintenance strategies. While the current model was trained under ideal dry conditions, the team plans to incorporate environmental variables, such as water film thickness and dust coverage in future iterations. Further validation across diverse climatic zones and pavement materials will enhance its global applicability.
This paper "XGBoost-based intelligent framework for asphalt pavement skid resistance assessment under different variables" was published in Smart Construction (ISSN: 2960-2033), a peer-reviewed open access journal dedicated to original research articles, communications, reviews, perspectives, reports, and commentaries across all areas of intelligent construction, operation, and maintenance, covering both fundamental research and engineering applications. The journal is now indexed in Scopus, and article submission is completely free of charge until 2026.
Citation:
Zhao Y, Hasan MRM, Zhang K, You L, Jamshidi A, et al. XGBoost-based intelligent framework for asphalt pavement skid resistance assessment under different variables. Smart Constr. 2025(4):0029, https://doi.org/10.55092/sc20250029.
