RUM and NON-RUM Path Choice Modelling for National Freight Transport



RUM and NON-RUM Path Choice Modelling for National Freight Transport

Authors

A Vitetta, A Quattrone, Mediterranea University of Reggio Calabria, IT

Description

In this paper a specification, calibration and validation of a path choice behavioural model for road freight transport at national level is proposed.

Abstract

The freight transport plays a fundamental role for the economy of every country. The greater part of goods is transported on road. In literature, few path choice behavioural models are specified, calibrated and validated for road freight transport.
In this paper a specification, calibration and validation of a path choice behavioural model is proposed. The path choice model is simulated in two phases:
1. choice set generation, that is the possible alternatives;
2. path choice among alternatives included in the choice set.
The choice set generation is realized with multi-criteria path generation.
The path choice is simulated with random (RUM) and non-random (no-RUM) utility models.
This paper represent an update of a previous paper presented in the ETC-European Transport Conference 2006 held in Strasbourg.
In particular, as regard the RUM, in addition of the previous models (Multinomial Logit and a C-Logit), in this new version of the paper is specified a Path-Size model.
Regarding the no-RUM, the updating involved the specification of a Fuzzy model. In the Fuzzy model it is considered the path? systematic disutility as the core of a fuzzy number. The disutility is calculated, in analogous way of the RUM, in function of the attributes through some coefficients (that must be calibrated). In this case for every alternative is calculated the possibility (not the RUM probability) that its disutility is the lowest. In order to make this, is estimated (using the maximum operator between two or more fuzzy numbers) the fuzzy number that represent the minimum disutility of all the alternatives, afterwards for every alternative is estimated the possibility that the disutility of such alternative is equal to the minimum previously calculated. Such possibilities must be transformed in choice percentages, adopting an opportune criterion of transformation of the possibilities distribution in one correspondent choice percentages distribution.
Path choice models were specified and calibrated on some test networks and also on a truck-drivers road-side survey. The chosen path was indicated in the questionnaire through origin, destination and intermediate nodes. In order to get more information about users path choice behaviour, the vehicles of some truck-drivers will be equipped with on board Intelligent Transportation Systems (ITS) for monitoring good transport. The utilization of ITS supplies new real time data about path choice. In all more of 200 interviews, with road-side survey and on board ITS, were kept for path generation and path choice modelling. Computations of path generation and level-of service-attributes computation were carried out using the national Italian road network which consists of all the motorways and the main national roads.
Relatively to path generation, the covered function is calibrated with the maximization of overlapping factor of chosen and generated paths. Relatively to path choice modelling, the calibration is carried out with the Maximum Likelihood method.
For the validation formal and informal tests on parameters and statistics about model goodness of fit have been carried out.
The experiment carried out on the Italian network showed that a significant degree of coverage of chosen paths could be obtained by generating paths with respect the criteria identified.

Publisher

Association for European Transport