Development of an accident risk prediction approach for dynamic route guidance.

摘要

Dynamic route guidance systems (DRG) aid drivers in choosing the best routes based on real-time conditions in networks. Whether for the evaluation of DRG's impact on a whole-network traffic safety or for determination of the safest routes by DRG, suitable accident prediction models are required. However, such models are not available for links of all kinds of roads. The objective of this research is to develop an accident prediction approach for links on freeways and urban streets suitable for DRG. Firstly, datasets were established by collecting accident data from police reports, link geometric and traffic data from a simulation network. Then, based on a concept we call “link neighbors”, a model form was set up. The model performance was measured using the standard deviation of the model's outputs, and through model optimization, model parameters were determined.