Modeling Driver Stop/Go Decisions in the Dilemma Zone During Amber Signal Phase Under Mixed Traffic Conditions

Abstract

Traffic safety and operational efficiency at signalized intersections are mainly influenced by drivers’ decision during the onset of the amber phase. The region in which drivers are uncertain whether to stop or proceed is known as the dilemma zone which poses significant risks of rear end and right-angle collisions. This study investigates factors affecting drivers’ stop/go decisions during the amber interval and develops a decision model for mixed traffic conditions. Two high‑crash signalized intersections, Sallaghari and Naya Thimi on the Kathmandu–Bhaktapur Road, were selected based on 2024 crash records. A binary logit model was estimated using variables extracted from videographic surveys and field observations. Video data were processed frame‑by‑frame at 30 fps using Python libraries such as OpenCV and YOLOv11. A total number of 6462 observations were recorded from both the intersections. Model results showed that approach speed, vehicle type (bus), distance to stop line, intersection type, and acceleration/deceleration were statistically significant in explaining the choice of a “Go” over a “Stop” decision. Overall, 57.97% of the drivers stopped before the stop line, 20.44% crossed during the amber signal phase and 21.59% were spotted in RLR. Buses were more likely to choose the “Go” decision, reflecting aggressive driving and competition for passengers. Accelerating drivers tended to proceed, while decelerating drivers were more likely to stop. The outcomes of this study can support the design of advance warning and enforcement strategies to alert drivers in advance and reduce crashes at signalized intersections.