Preempt or yield? An analysis of driver's dynamic decision making at unsignalized intersections by classification tree

For developing countries and regions, due to less construction of stop signs and roundabouts, as well as limited regulation of driving courtesy, safety issues at unsignalized intersections require harder concern. In China, drivers rarely stop completely at unsignalized intersections, but gradually enter and dynamically make their decisions to yield or preempt by gaming with other vehicles. Wrong decisions made in this quick process often lead to accidents. In this study, we aimed to explore how straight drivers dynamically made decisions when encountered merging vehicles at unsignalized intersections in China. By video graphing traffic scenarios, 150 cases of merging traffic were selected at a 4-legged unsignalized intersection in Kunming City. Motion parameters of the vehicles were extracted from video detection software. By modeling the motion parameters to a classification tree, the decision moment of straight drivers' yielding/preemptive decision and the motion parameters which influenced drivers' decision significantly were identified. Results showed that straight drivers made yielding/preemptive decisions 1.3-1.5 s before reaching the merging point. Speed difference between the straight vehicle and the turning vehicle was the most important factor to impact straight driver's decision-making. Turning vehicle's speed and distance to the merging point also impacted straight driver's decision. Moreover, a U-shape curve was found when plotted the minimum gap between the two vehicles by the speed difference of the two vehicles at the decision moment (1.3 s). The accurate motion parameters found in this study helped to develop driver's thorough behavior model at unsignalized intersections, and suggest safety measures further.