Performance Comparison Between Artificial Neural Network and Analytical Models for Real-Time Cycle Length Design

A searching-free analytical procedure is mostly utilized with Adaptive Traffic Control Systems (ATCS) because of its short and stable computation time even though it suffers from feasibility limited to the designed conditions. Artificial Neural Network (ANN) is a perceptual model, which may overcome the shortcomings of an analytical model. However, ANN was concerned about its sensitivity that may cause safety problems in the field. This paper presents performance comparison between an analytical model, a degree-of-saturation based cycle length design model of Cycle-Offset-Split-Models-Of-Seoul (COSMOS) and an ANN model developed for this study. Cycle lengths from these models were compared against the ones suggested by TRANSYT-7F and SYNCHRO at various demand levels. It was found that the ANN model provides with the optimal cycle lengths stably adjusted by the minimum, the maximum, and a cycle increment, while the analytical model promotes congestion at certain operational conditions considered in the test.

Language

  • English

Media Info

  • Media Type: CD-ROM
  • Features: Figures; References; Tables;
  • Pagination: 22p
  • Monograph Title: TRB 85th Annual Meeting Compendium of Papers CD-ROM

Subject/Index Terms

Filing Info

  • Accession Number: 01025971
  • Record Type: Publication
  • Report/Paper Numbers: 06-2038
  • Files: TRIS, TRB
  • Created Date: Jun 28 2006 8:47AM