Data-Driven Surrogate Models for Road Maintenance Scheduling

A recent study published on arXiv introduces data-driven surrogate models aimed at estimating equilibrium traffic flows for planning road maintenance at the network level. Utilizing traffic data from Newark, New Jersey, the research tackles the difficulties associated with repeatedly resolving equilibrium traffic assignment models. The authors create surrogate models that predict arc flows based on origin-destination demand, using optimization-derived equilibrium solutions as a benchmark. This method is proposed as a scalable component for future maintenance scheduling systems. Classified under Electrical Engineering and Systems Science, particularly Systems and Control, the paper was submitted on August 26, 2025 (arXiv ID 2608.14491). The findings underscore the promise of machine learning in improving infrastructure management by minimizing computational demands while ensuring precision.