Using electric vehicle onboard data for pavement quality assessment and management

Summary

This project investigated the feasibility of using electric vehicle onboard diagnostics (OBD-II) data for pavement quality assessment in Minnesota, inspired by Denmark's Live Road Assessment (LiRA) project. The research addressed the critical need for cost-effective, continuous pavement monitoring to supplement expensive instrumented vehicle surveys traditionally used by the Minnesota Department of Transportation (MnDOT). The methodology involved a comprehensive OBD-II scanner evaluation, selecting the OBDLink LX device for its superior compatibility with electric vehicles and ability to collect 694 Parameter IDs (PIDs). Data collection occurred across three test routes: MN-36 loop, Metro to Northfield loop, and MnROAD facility, generating around 10,000 data points synchronized with MnDOT's high-precision Pathways van measurements. The LiRA software framework was adapted through systematic data preprocessing, feature engineering using Sequential Feature Selection, and machine learning model development. Results demonstrated that hybrid Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) models achieved the best performance when using speed combined with vehicle orientation and acceleration features, yielding root mean square error (RMSE) values of 0.172 compared to MnDOT reference data. The methodology showed correlation with both MnDOT measurements and NIRA Dynamics network-level data across diverse road conditions, though performance varied based on environmental factors and data collection conditions. Implementation guidelines were developed for large-scale deployment, including equipment standardization protocols, data-processing pipelines, and risk-mitigation strategies. The research indicates that vehicle sensor data can provide useful pavement condition information for continuous infrastructure monitoring with reduced incremental cost. This approach shows potential for supplementing traditional pavement management systems, though successful implementation requires careful consideration of multiple operational factors and validation protocols to ensure reliable performance across varying conditions.