An application of Wavelets and Self-organizing Kohonen Maps to Identify Transitions in Short-Term Urban Traffic Flow

Most short-term forecasting efforts make the simplifying assumption that traffic and its evolution are a continuous process that can be modeled by time series approaches. However, shifts to extreme values (congestion) and the irregular nature of series of traffic variables suggest the lack of continuity in traffic flow series. The present paper aims at identifying the various traffic flow conditions with respect to the transitional behavior of traffic flow. Transitions are detected by isolating the local maxima in the dyadic wavelet transformation of volume and occupancy series. Traffic flow conditions are identified using a 2-stage self-organizing neural network approach. The implementation on a congested signalized arterial results in identifying the boundaries of four distinct areas of traffic flow. The approach succeeds in numerically define the onset of congestion. Finally, the complexity in synchronized flow is reflected by the occurrence of two distinct areas of traffic flow.

Language

  • English

Media Info

  • Media Type: CD-ROM
  • Features: Figures; References; Tables;
  • Pagination: 22p
  • Monograph Title: TRB 86th Annual Meeting Compendium of Papers CD-ROM

Subject/Index Terms

Filing Info

  • Accession Number: 01049354
  • Record Type: Publication
  • Report/Paper Numbers: 07-0272
  • Files: TRIS, TRB
  • Created Date: Feb 8 2007 4:49PM