The objective of this paper is to present and validate a closed-loop framework for the evaluation of the impact of predictive route guidance in case of severe incidents. The presented framework has been implemented through the integration of two state-of-the-art tools: the microscopic traffic simulation laboratory MITSIMLab (which represents the "ground-truth" real network conditions) and the dynamic traffic assignment system DynaMIT (which acts as the Traffic Management Center and generates predictive and consistent dynamic route guidance). Drivers can thus make informed decisions about their route choice when they receive information through Variable Message Signs (VMS). The impact of dynamic route guidance was evaluated using the developed closed-loop framework in the lower Westchester County network just north of New York City. The large-scale nature of the application provides a unique environment for the real-world validation of the presented framework. Two alternative scenarios were modeled, each involving one hour-long incident with full blockage on all lanes for the peak direction. The selected incidents are representative of the frequent closures and bottlenecks that occur on the network, both in terms of severity and location. The sensitivity of the effectiveness of the generated route guidance with respect to the number of prediction iterations used was assessed, and it was found that in this application considerable benefits were obtained even when a single prediction iteration was performed. The marginal benefits for using more than three prediction iterations were found to be limited, indicating that, in this case, three prediction iterations provide a practical trade-off between maximum effectiveness and computational feasibility.
Abstract