Feasibility of using autonomous vehicle data for road rutting estimation

Pavement rutting, a common surface deformation occurring in wheel paths due to traffic loading, poses risks to driving safety and pavement integrity if not effectively managed. Autonomous vehicles (AVs) are experiencing rapid development thanks to advancements in perception sensors. Real-time pavement rut estimation could enable AVs to adjust their lateral position to avoid existing ruts, thereby reducing passenger risk and limiting further pavement deterioration. This study investigated AV-based estimation of rutting depth and cross-sectional area, using data from the Autonomous Research Vehicle Observatory (ARVO), equipped with a Velodyne VLS-128 LiDAR sensor (Distance of Adjacent Points (DAP): 2.8–4.6 cm). For accuracy assessment, reference data were provided by the Roamer-R4DW system (Roamer; DAP: 4–6 mm), a mobile mapping system equipped with a Riegl VUX-1HA scanner. Roamer’s reliability was further validated against high-precision data from the Pavement Profile Scanner (PPS; DAP: 3–5 mm). Field experiments were conducted along a 3.5 km road section exhibiting severe rutting, and an enhanced rut measurement algorithm was applied to accurately extract rut features. Through the study, ARVO achieved root mean square errors (RMSEs) of 2.8 mm for rut depth and 21.3 cm 2 for cross-sectional area, with corresponding biases of –2.2 mm and –2.2 cm 2 . Roamer’s accuracy, validated against PPS, yielded estimation error (RMSE) and bias of 1.2 mm and 0.3 mm for depth, and of 13.3 cm 2 and 10.4 cm 2 for cross-sectional area, respectively. Results demonstrate that AV-based data can accurately estimate pavement rutting, supporting both roadway maintenance and driving assistance.