Generalised Maximum Entropy Methods in Transportation Planning



Generalised Maximum Entropy Methods in Transportation Planning

Authors

LEURENT F M, INRETS, France

Description

The maximum entropy (ME) estimation of an origin-destination matrix has become a

Abstract

The maximum entropy (ME) estimation of an origin-destination matrix has become a
classical transportation engineering tool. At first proposed by Wilson (1967) to
determine the matrix with the highest probability of satisfying a set of origin- or
destination-based eoustraints, it ha~ been successfully extended to make use of traffic
counts (e.g. Van Zuylen and Willumsen, 1980). Recently Golan et ai. (1996) have
generalized the ME principle to address a broad range of estimation problems,
including ill-conditioned lInear inverse problems with noise, simultaneous equations,
linear and non-linear dynamic systems and discrete choice problems.

The paper aims to examine the possible eontributious of the generaiized ME (GME)
method to transportation planning problems. After a review of the ME principle, the
GME method is inlrodueed and exemplified; an extended framework is designed for
O-D matrix estimation. Then we discuss the application of the GME method to
discrete choice models and we put forward a novel GME estimator which lakes
account of cousumer preferences. Finally, the GME method is assessed and research
directions are described.

Publisher

Association for European Transport