Modelagem do impacto das motocicletas na fluidez do tráfego de interseções semaforizadas de Fortaleza

Abstract

Despite the significant growth in motorcycle usage in Brazilian cities and several other developing countries, traffic models still lack proper representation of this transport mode. One of the main challenges lies in modeling motorcycle infiltration behavior in queues at signalized intersections and its effects on traffic flow. This research aims to model the impacts of motorcycles on traffic flow at signalized intersections in Fortaleza, Brazil. Three specific objectives were pursued: (i) to propose an automated vehicle trajectory data collection procedure using computer vision; (ii) to define variables capable of representing the effects of motorcycles on signalized intersection traffic flow; and (iii) to estimate and analyze such effects based on real data. Data collection was carried out using drone-based aerial footage, applying the YOLOv10 object detector and the StrongSORT tracking algorithm. Automatic correction mechanisms were implemented to compensate for camera displacements due to flight instability. Case studies included intersections on two arterial roads in Fortaleza: Av. Carapinima (with and without exclusive motorcycle waiting areas, or ―motoboxes‖) and Av. Bezerra de Menezes (with motobox). The computer vision tool showed satisfactory performance for the most representative and visually distinct vehicle classes (cars, motorcycles, and trucks), and more limited accuracy for buses, vans, bicycles, and pedestrians. Validation against manual data collection using the RUBA software indicated high accuracy, with negligible differences (≤ 0.1 s) in vehicle passing times and headways. In the statistical modeling of headways for the first vehicles in the queue, the presence of motorcycles ahead of the first vehicle and between vehicles longitudinally was statistically significant, whereas motorcycles in the virtual corridor showed null or minimal impact. The estimated passenger car equivalents (PCEs) for motorcycles ranged from 0.07 to 0.09 motorcycles/PCE, reflecting the average effect of all motorcycles crossing the stop line, including those with no direct interaction with other vehicles — thus explaining the lower values compared to the literature. Scientifically, this study advances the modeling of multimodal urban traffic, particularly in contexts with high motorcycle volumes. From a practical standpoint, the findings support capacity and level of service analyses at signalized intersections and can enhance traditional traffic models, enabling more realistic and reliable analysis for decision-making at strategic, tactical, and operational levels.