Proactive crash risk prediction modeling for merging assistance system at interchange merging areas

Abstract Objective: Ramp drivers have to merge into the through traffic in a limited time and space at interchange merging areas. Different merging decisions are made due to drivers’ various perception abilities of potential danger, which might significantly increase the crash risk. Driving assistance technology (DA) is expected to be an effective way of mitigating the crash risk. Hence, this paper aims to contribute to the literature by designing a model strategy to predict the crash risk of merging drivers in order to enhance the merging assistance system for crash avoidance. Methods: Unmanned aerial vehicle (UAV) was used to collect individual vehicle data to conduct traffic analysis at the microscopic level. A model strategy was proposed to predict the crash risk of merging vehicles which could make sure that ramp drivers are aware of potential risks in advance. Three models (i.e., binary logistic regression, multinomial logistic regression, and nested logit models) were developed and compared. Results: Target-lane-related and merging-vehicle-related variables were found significant with crash risk, including the speed of the merging vehicle, the speed of lead/lag vehicle in the target lane, the type of lead/lag vehicle in the target lane. Different variables were found to be significant in the proposed models. Conclusions: The results suggest that the nested logit model has the highest prediction accuracy. It is concluded that the merging speed, driving ability (i.e., lane-keeping instability), and the vehicle type in the target lane affect the crash risk. Finally, the implementation of the proposed prediction model for merging assistance system is designed. The findings from this study can have implications for the design of the merging assistance system for helping drivers make safe merging decisions and thus enhancing the safety of the interchange merging area.