Efficient Metamodel Framework for Nonlinear OD Matrix Estimation Problem

The last few decades have produced a rich body of literature on dynamic origin and destination (OD) matrix estimation devoted to the development of heuristic methods for reducing the high computational effort of traffic simulation with a low trade-off in the accuracy of the results. This paper proposes a metamodel as a function that maps the nonlinear relationship between traffic observations and OD flows. The derived metamodel framework is consistent with all of the estimated demands through iteration steps, which results in a non-constant, demand-dependent assignment matrix. Furthermore, the metamodel requires the same amount of traffic simulations as the traditional approach, to maintain a low computational cost. In addition, a stochastic gradient method, which escapes from local minima more efficiently, is applied to solve the dynamic OD estimation problem. The performance of the proposed metamodel and solution approach has been evaluated for the formulation of four OD estimation methods. The least square OD matrix estimation model solved with gradient descent algorithm has been selected as a complementary method, widely used in practice, whose limitations we aimed to solve. Numerical experiments are performed on Vitoria network, Spain, with real traffic count data and two historical OD patterns to evaluate the performance of the proposed solution approach. The results show that the proposed stochastic gradient metamodel method demonstrates a stable decrease in the cost function with the lowest value throughout the iteration steps. The experiments indicate its robust performances in terms of quality in both of the initial OD matrices, particularly when the network is congested.

  • Supplemental Notes:
    • This paper was sponsored by TRB committee ADB30 Standing Committee on Transportation Network Modeling.
  • Corporate Authors:

    Transportation Research Board

  • Authors:
    • Djukic, Tamara
    • Masip, David
    • Breen, Martijn
    • Casas, Jordi
  • Conference:
  • Date: 2019


  • English

Media Info

  • Media Type: Digital/other
  • Features: Figures; Maps; References; Tables;
  • Pagination: 14p

Subject/Index Terms

Filing Info

  • Accession Number: 01698266
  • Record Type: Publication
  • Report/Paper Numbers: 19-05188
  • Files: TRIS, TRB, ATRI
  • Created Date: Mar 1 2019 3:51PM