The Impact of Various Uncertain Information Schemes On Route Choice



The Impact of Various Uncertain Information Schemes On Route Choice

Authors

E A I Bogers, S P Hoogendoorn, Delft University of Technology, NL

Description

Abstract

This paper describes an experiment that gives insight into the roles of inaccurate (i.e. incomplete and/or unreliable) information in route and departure time choice. This choice is highly influenced by how people learn and how they deal with uncertainty. Information is expected to reduce this uncertainty and to facilitate the learning process.
But, just like in real life, the information provided itself is inaccurate as well, because it is based on the route characteristics as they have been observed instead of as they will actually be at the time the traveller will enter the route.

Two major approaches are commonly used to observe travel behaviour: Stated Preference (SP) methods, and Revealed Preference (RP) methods. The interactive Travel Simulator Laboratory (TSL) is a hybrid approach, combining the benefits of both approaches, while avoiding their respective limitations. The TSL presents the traveller with a choice between two given routes. For both routes the free flow time is given. Just before he has to make a choice, uncertain information about queue lengths on both routes is provided via a drip. To capture the learning process, the traveller is asked to make this choice several times, representing several days. His choices are input to a traffic simulation model. This model computes the realised traffic flows and the resulting travel times. The model also accounts for peak periods. An important feature is that, on the contrary to SP surveys, the traveller is also confronted with the consequences of his decision during the experiment. A bad decision, i.e. a decision with a high realised travel time, results in a long waiting time before the next choice can be made.

Three information schemes are carried out. They all give pre-trip information on queue lengths on both routes. The first also gives ex-post information on the travel time of the chosen route, the second adds the travel time at the non-chosen route and in the third, most elaborate one, the realised travel time on both routes remains on-screen for all past periods. In order to find which information scheme works best, in terms of providing the traveller with the lowest disutility (or highest utility), and/or making him learn the fastest, multinomial logit models are estimated.

Results show that travellers learn significantly faster under the most elaborate information scenario. Furthermore, the number of times a traveller has already chosen a route increases the utility of that route significantly. Finally, the expected travel time of the route and the travel time, lateness and earliness at the most recent period all prove to be significant.

Publisher

Association for European Transport