The relationships between traffic flow variables play important roles in traffic engineering. They are used not only in basic traffic flow analyses but also in some macroscopic traffic flow simulation models. For many decades, various mathematical formulations that describe the relationships among density, flow, and speed have been proposed, including multiregime models. Previously, the best mathematical curve was determined by trying several different formulas and applying regression analysis. In these processes, one must specify in advance which mathematical formula should be adopted and where it should be shifted to another in a multiregime model. Neural network models have some promising abilities to represent nonlinear behaviors accurately and to self-organize automatically. A procedure for describing the macroscopic relationships among traffic flow variables using some neural network models is presented. First, a Kohonen feature map model was introduced to convert original observed data points into fewer, more uniformly distributed ones. This conversion improved regression precision and computational efficiency. Next, a multilayer neural network model was introduced to describe the two-dimensional relationships. The model was effective in describing the nonlinear and discontinuous characteristics among traffic flow variables. It was unnecessary to specify the regression curves and the transition points in advance. The multiple correlation coefficients resulting from the model were better than those resulting from a conventional nonlinear equation.


  • English

Media Info

  • Features: Figures; References; Tables;
  • Pagination: p. 11-18
  • Monograph Title: Traffic flow theory and characteristics with applications for intelligent transportation system technologies
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00715620
  • Record Type: Publication
  • ISBN: 0309062039
  • Files: TRIS, TRB
  • Created Date: Jan 19 1996 12:00AM