Random Covariance Heterogeneity in Discrete Choice Models



Random Covariance Heterogeneity in Discrete Choice Models

Authors

S Hess, Imperial College London and RAND Europe, UK; J Polak, Imperial College London, UK; D Bolduc, University of Laval ? Quebec City, CN

Description

This paper discusses the extension of the discrete choice framework to the analysis of random variations in the correlation structure across individuals.

Abstract

As our time schedules become tighter everyday, mobility and flexibility characterize our lives. Modern information technologies keep us up-to-date and make our mobility planning easier wherever we are: at home, in the office, or while travelling. Transportation companies have to respond to this trend. Meanwhile everybody knows, that passenger satisfaction and the attractiveness of transport services are increased considerably, if reliable and clear passenger information is available wherever needed.

Today, many systems are in use and the experience is twofolded: The task of providing the information is highly difficult and full of pitfalls. To provide a real real-time system, extensive experience is required which then results in trustworthy algorithms for analysing the traffic and transit situation and predicting suitable departure times at the stop, in the Internet and to the mobile phone.

A sophisticated forecasting algorithm has to determine the exact vehicle arrival and departure times at the stops. Many more parameters have to be considered than only the current schedule situation. For example, the vehicle-specific minimum turn and stopover times have to be considered in these calculations as route-specific catch-up possibilities or dynamic signal priority criteria. All of these influencing factors should be weighable individually and parameterisable so that the algorithm can take specific circumstances into account.

The system has to be adapted exactly to the situation and the requirements of the respective transportation company. How can existing equipment be integrated to save costs? And how can future system enhancements be considered already in the design and specification phase to create an system that is both capable and efficient?

The paper will discuss the difficulties with real-time passenger information and explain resolution strategies. It will also give examples of system implementations and high-light related implementation problems.

Publisher

Association for European Transport