Capacity-Aware Deep Learning for Generalizable Traffic Volume Estimation Across Links and Cities

2026-07-27

AI Summary

This study addresses the poor generalization of traffic flow estimation in sparsely monitored road networks caused by reliance on fixed sensors. To overcome this limitation, the authors propose a link-level deep learning approach that leverages readily available data—including probe vehicle speeds, road topology, and weather conditions. The method introduces capacity-aware modeling to decouple traffic flow into structural capacity and time-varying utilization, incorporating fundamental traffic flow theory as hard constraints. By integrating supervised local mapping learning with out-of-distribution spatial generalization training strategies, the model achieves significantly enhanced transferability across road segments and cities. Experimental results demonstrate that the proposed approach consistently outperforms state-of-the-art models under diverse cross-domain settings, confirming the effectiveness of structural constraints in mitigating spatial distribution shifts.

Abstract

Network-wide traffic volume estimation typically relies on propagating measurements from fixed sensors, making performance highly dependent on sensor density and limiting deployment in sparsely instrumented networks. We propose a link-level learning framework that estimates hourly traffic volumes from widely available territorial data only, including probe speed profiles, road and topological descriptors, along with weather observations. A supervised local mapping is learned from sparse sensor measurements and evaluated under two generalization settings: intra-network (unseen links within the training network) and inter-network (unseen city). This formulation frames traffic volume estimation as a spatial out-of-distribution generalization problem under sparse supervision. To enhance spatial robustness, we introduce a capacity-aware formulation that models volume as the product of a link-specific structural capacity and an hourly regime-aware utilization ratio, embedding traffic-theoretic constraints directly into the learning process. Extensive experiments in both generalization settings demonstrate that the proposed structural constraints consistently outperform a state-of-the-art baseline under spatial distribution shift.

Problem

• traffic volume estimation
• spatial generalization
• sparse supervision
• distribution shift
• network-wide inference

Innovation

• capacity-aware learning
• traffic volume estimation
• spatial generalization
• structural constraints
• out-of-distribution robustness