Calibrating Dynamic Network Models for Traffic Operator Decision Support

To support traffic operators in regional traffic management centers with their network control tasks, the Traffic Research Center of the Dutch Ministry of Transportation has proposed a two-step scenario-based approach. In the first step, traffic engineers prepare candidate traffic control scenarios that are likely to resolve problems that occur in real-world network traffic operations. In the second step, network operators can try out a limited number of control scenarios with an online prediction system to see which of these best resolves the problems at hand. To prepare the scenarios, a scenario assessment system is used that predicts the impacts of a specific control scenario by using a dynamic network traffic flow model (MetaNet). The development of a fully automated calibration approach for the input and parameters of this macroscopic network simulation model is described. The approach presented consisted of four steps—preparing inflows and turning proportions, establishing lane-specific traffic flow parameters, estimating the global simulation parameters, and determining unresolved model inputs—eventually leading to input data and a set of parameters that yield the best prediction of networkwide traffic conditions. Furthermore, the approach checks the available input data for inconsistencies in terms of network configuration and measurement errors. On the basis of the application examples tested, the approach is well suited for the calibration task at hand, yielding a remarkable improvement in the prediction error. Furthermore, the computation time for a realistic network is sufficient for most practical applications.

Language

  • English

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 01049339
  • Record Type: Publication
  • ISBN: 9780309113038
  • Files: TRIS, TRB, ATRI
  • Created Date: Feb 8 2007 5:25PM