A Random Parameter Logit Model of Red Light Violation of Cyclists at Urban Overlap Phase Signal Intersections

This study aims to identify factors influencing cyclists’ red-light violations (RLV) at overlap phase signal intersections. Research data containing 10,733 cyclists’ crossing behavior came from a field survey of seven intersections that use overlap phases signal control schemes in Guangzhou, China. A new method of dividing the signal cycle according to the different risk sources of cyclists is proposed, namely, the reverse green-light risk period, the concurrent green-light safety period, the concurrent green-light risk period, and the vertical direction risk period. Afterward, the random parameters logit model (RPLM) was developed to study the impact of the cyclists’ features, arrival time, and intersection environment. The study’s results confirm the superiority of the random parameter model fit that considers cyclists’ heterogeneity. In addition, males, professional cyclists, hauling, end-of-red-light period arrival, concurrent green-light risk period arrival, without helmets, induce RLV behavior. Among these factors, hauling, end-of-red-light period arrival, and concurrent green-light risk period arrival are random variables. In the new signal cycle classification method, cyclists arriving during the concurrent and reverse green-light risk periods tend to run red lights. Based on these conclusions, specific and insightful strategies are proposed regarding policy management and intersection signal timing optimization to improve safety levels at overlapping phase signal intersections.