Domain Adaptation Framework for Turning Movement Count Estimation with Limited Data

Abstract

Accurate turning movement counts (TMCs) at intersections are crucial for traffic signal control, congestion mitigation, and road safety. Recent advancements in machine learning and data-driven approaches have offered promising alternatives for estimating TMCs. However, traffic patterns can vary significantly across different intersections because of factors such as road geometry, traffic signal settings, and local driver behaviors. This domain discrepancy limits the generalizability and accuracy of machine learning models when applied to unseen intersections. To address these limitations, this research proposes a novel framework leveraging domain adaptation (DA) to estimate TMCs at intersections by using traffic-controller-event-based data, road infrastructure data, and point-of-interest data. Evaluated on 30 intersections in Tucson, Arizona, U.S., the proposed DA framework consistently outperformed state-of-the-art models, achieving the lowest error values across all movements. The framework obtained mean absolute error values of 12.29 for left turns, 34.39 for through movements, and 16.17 for right turns. A similar pattern was observed for root mean squared error, with the model achieving 15.73 for left turns, 43.90 for through movements, and 20.08 for right turns, outperforming all comparison methods.