Towards sustainable and efficient data collection using dynamic sampling for gravel road condition assessment

Artikel i vetenskaplig tidskrift, 2026

This paper examines a dynamic sampling technique for optimising the collection and storage of gravel road condition data by reducing redundancy while preserving accuracy. Advances in sensor technology and Information and Communication Technology (ICT) enable large-scale condition assessment but generate vast amounts of data, making efficient collection and storage essential. The Dynamic Sampling Rate Algorithm–Parametric Machine Learning Optimisation (DSRA-PMLO) was applied to simulated datasets and vehicle vibration response (VVR) signals collected using an Integrated Electronics Piezo-Electric (IEPE) accelerometer on three gravel roads with varying surface conditions. The algorithm was evaluated using error thresholds of 5–50%. Results show that DSRA-PMLO substantially reduces data volume without significant information loss. At a 20% error threshold, approximately 50% of the original data was sufficient for accurate condition assessment, while faults including potholes, corrugation, and loose gravel remained reliably detectable. Reduced data collection also lowers storage, bandwidth, and energy requirements.