講演情報

[Data_Collection-35]Development of a Pothole and Pavement Distress Detection Model Using Vehicle Big Data

*Yuki Ito1, Kyeonghwan Kaku1, Shinpei Nomoto1, Arata Sakurai2, Takafumi Kurano2 (1. Maeda Road Construction Co.,Ltd (Japan), 2. Toyota Motor Corporation)

キーワード:

big data、road surface surveys、potholes、pavement damege、machine learning

Maintaining pavement surfaces efficiently is a major challenge for municipalities due to limited budgets and workforce shortages. Traditional inspection methods, such as visual surveys and pavement condition measurements, are labor-intensive and costly, while simplified techniques using smartphones or vehicle-mounted sensors still require regular patrols. To address these issues, this study proposes a pothole detection model utilizing vehicle big data collected from connected cars. The model incorporates multiple features derived from wheel speed variations, including average and maximum fluctuations, and applies machine learning techniques to identify pavement distresses (e.g., potholes and localized depressions) requiring urgent repair. Preliminary experiments on a dedicated test track with artificial potholes confirmed that damage exceeding tire width and asphalt layer depth significantly influences sensor readings, enabling accurate detection. The proposed model was applied to municipal roads over an 11-week period. Results show that non-critical locations can be excluded after a certain observation period, while areas requiring repair achieved a precision of approximately 0.4-0.6 across all periods. Additionally, candidate locations likely to deteriorate in the future were successfully identified, supporting proactive maintenance strategies. This approach demonstrates the feasibility of integrating vehicle big data into pavement management systems, reducing patrol workload by up to 30% and annual maintenance costs by approximately 10%. The findings highlight the potential of big data-driven solutions to shift road management from reactive to preventive maintenance, contributing to sustainable infrastructure management.