ARTIFICIAL NEURAL NETWORKS IN THE TRANSPORTS

LE RETI NEURALI ARTIFICIALI NEI TRASPORTI

ARTIFICIAL NEURAL NETWORK (ANN) ARE A MATHEMATICAL TOOL BELONGING TO LEARNING SYSTEMS: THEY CAN BE ADAPTED AND LEARNED WITH A CONSIDERABLE FLEXIBILITY AND EFFICIENCY, AND FURTHERMORE PROCESSING TIMES ARE GREATLY REDUCED. THE TYPOLOGIES OF ARCHITECTURE WE CAN FIND IN SCIENTIFIC WORKS IS VERY BROAD AND THEY ALLOW US TO IMPLEMENT CLASSIFICATION, OPTIMIZATION AND RECOGNITION FUNCTIONS. THE THREE BEST KNOWN AND DEVELOPED TYPOLOGIES, IN THEORY AND WORKING APPLICATIONS ARE: BACKPROPAGATION, HOPFIELD AND KOHONEN. OF THESE THREE, BACKPROPAGATION IS THE MOST WIDELY UTILIZED AND HAS BEEN APPLIED IN TRANSPORTATION WORKS IN THE FOLLOWING FIELDS: 0/D MATRIX IDENTIFICATION; INCIDENT DETECTION; PREDICTION OF VEHICULAR FLOW PARAMETERS; SIMULATION OF USER BEHAVOUR; RECOGNITION OF FLOW PARAMETERS FROM VIDEO IMAGES OF FLOW. OTHER APPLICATIONS ARE EITHER CURRENTLY BEING STUDYED OR ARE OF EASY APPLICATION: FOR EXAMPLE FOR THOSE PROBLEMS BASED ON THE OPTIMIZATION OF A FUNCTION. A FIRST THEORETICAL APPLICATION OF ANN ON UNINTERRUPTED FLOW CONTROL IS PROPOSED. THIS ANN IS CAPABLE OF RECOGNIZING TRAFFIC CONDITION AND FORECASTING ITS STABILITY OR INSTABILITY IN SPACE AND TIME.

  • Availability:
  • Corporate Authors:

    N/A

    ,   United States 
  • Authors:
    • Mussone, L
  • Publication Date: 1994-6

Language

  • Italian

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 00715741
  • Record Type: Publication
  • Source Agency: TRL
  • Files: ITRD
  • Created Date: Jan 31 1996 12:00AM