Physics-guided estimation of freight vehicle loading status using digital tachograph data.

Accurately distinguishing between empty and loaded freight vehicle operations is essential for improving logistics efficiency, reducing environmental impacts, and supporting evidence-based freight and infrastructure policies. In Korea, where empty running accounts for over 40% of truck trips, reliable load-status information can enable demand-driven routing, enhance freight origin–destination (OD) estimation, strengthen overloading enforcement, and optimize road maintenance strategies. This study proposes a scalable framework for load-status estimation using Digital Tachograph (DTG) data, a legally mandated system that provides high-frequency records of speed, acceleration, and engine RPM without requiring additional sensors. Physics-informed features—derived from vehicle dynamics, drivetrain behavior, and resistance forces—were constructed and used in a Bayesian Neural Network (BNN) classifier to incorporate both predictive accuracy and uncertainty quantification. Empirical results show that the proposed method achieves an average accuracy of 85.3% when using 9-s averages, exceeding 90% at highway speeds, and approaching full accuracy when temporal aggregation is applied. These findings demonstrate not only the technical feasibility of DTG-based estimation but also its capacity for nationwide, real-time monitoring of freight operations. Beyond model performance, the results highlight the broader policy relevance of this approach. By leveraging existing DTG infrastructure, the method offers a cost-effective and field-deployable solution for enhancing freight system visibility, reducing empty running, improving sustainability, supporting overloading detection, and informing infrastructure management. This positions DTG-based load-status estimation as both a methodological contribution to transportation research and a strategic decision-support tool for policymakers and industry stakeholders.