Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!udel!gatech!swrinde!pipex!sunsite.doc.ic.ac.uk!cc.ic.ac.uk!amoss@ic.ac.uk
From: amoss@cvcgi.ic.ac.uk (Amos Storkey)
Subject: Neural Transportation Models
Message-ID: <1994Nov21.164445.24619@cc.ic.ac.uk>
Originator: amoss@cvcgi
Keywords: Transport Modelling Hopfield
Sender: amoss@cvcgi (Mr Amos James Storkey)
Nntp-Posting-Host: cvcgi.cv
Organization: Imperial College  of Science, Technology and Medicine, London
Date: Mon, 21 Nov 94 16:44:44 GMT
Lines: 31





Does anyone know of any work has gone on in looking at applications of Neural Networks
to Transport Modelling or Traffic Modelling. I am aware that Neural Technologies are currently doing
work for AGD Systems putting some Neural Intelligence onto systems of sensors already developed
by AGD, and which would be in traffic congestion analysis (esp looking at the effect of different
vehicle sizes). However this project is at an early stage of feasibility assessment, and confidentiality
agreements make it hard to squeeze any information out.

Clearly the work done with Neural Networks (esp Hopfield Nets) on specific problems like the travelling
salesman problem, are related to areas of transport modelling. Transport Models have historically
rarely moved beyond the dubious benefits of regression analysis, and have been a rather inexact
science. It appears to me that General Regression Neural Networks would have something to offer
as a starting point. Also the easy representation of transport networks as weighted digraphs would
surely make them look favourable to types of Neural Modelling Analysis.

I am interested in anyone who:

a) Knows of work done on any of these areas
or
b) Knows of somewhere interested in supervising a Phd research project in this area, which
might attract suitable funding.



Amos Storkey
London University Centre for Transport Studies
Imperial College		University College
London University
