Decision-Making Tool for Road Preventive Maintenance Using Vehicle Vibration Data

Ahn, Changbum; Wang, Chao; Du, Jing · 2019 · ROSA P / Transportation Consortium of South-Central States

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This study addresses the challenge of optimizing road preventive maintenance in the face of aging infrastructure and limited funding. Current inspection methods are either subjective and labor-intensive (manual inspection) or accurate but prohibitively expensive for widespread deployment (LiDAR, ground penetration radar). The research proposes a crowdsourcing approach that leverages vehicle vibration data collected via smartphone sensors to detect specific road damages, such as potholes and cracks, enabling cost-effective, high-frequency monitoring. The methodology involved developing a cloud-based smartphone application to collect real-time vehicle vibration data, location information, and road damage imagery from drivers in Texas and Louisiana. A total of 310 miles of road-induced vehicle vibration data was gathered. To address the high noise levels inherent in crowdsourced data—caused by varying vehicle types, conditions, and driving behaviors—the researchers employed deep learning techniques, specifically Sparse Coding and Self-Taught Learning (STL). The study compared these deep learning methods against traditional machine learning classifiers, including Support Vector Machines (SVM), Decision Trees, and Multilayer Perceptrons (MLP). Data preprocessing included reorienting acceleration data to a global frame of reference and filtering signals, while data augmentation techniques were applied to enhance model training. The results demonstrated that traditional classification methods struggled with the noisy, low-quality nature of the crowdsourced vibration data. In contrast, the deep learning approach, particularly the features generated through Sparse Coding, significantly enhanced detection performance. Sparse Coding allowed the model to identify fundamental elements ("atoms") within the vibration signals, effectively filtering out noise and reconstructing signals to identify specific damage types. The STL algorithm proved effective in handling the limited labeled data available for training. The developed system successfully classified diverse road damages, moving beyond simple roughness indices (IRI) to identify specific distress types. The significance of this work lies in its potential to transform infrastructure maintenance from an ad hoc process to a data-driven paradigm. By utilizing ubiquitous smartphone sensors, the proposed tool offers an economical, scalable solution for continuous road condition monitoring. This enables decision-makers to build precise temporal deterioration models, facilitating proactive preventive maintenance that preserves road functionality at minimal cost. The study highlights the viability of citizen science in engineering, leveraging the collective data of regular drivers to solve complex infrastructure challenges that automated sensors alone cannot address efficiently.

Key finding

Sparse Coding features significantly enhanced road damage detection performance by addressing the low-quality issues inherent in crowdsourced vehicle vibration data.

Methodology

field_study

Sample size: 310

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• telematics crash prediction