Modelling the Impacts of ATIS Accuracy on Travellers' Route-choice Behaviour and Risk Perception
E Ben-Elia, University of the West of England, UK; B Gennaro, D Pace Roberta, University of Naples, IT; I Frederico, Y Shiftan, Israel Institute of Technology, IL
The impact of the level of accuracy of information provided to travellers is investigated using a discret-choice modelling approach in an experimental setting.
ATIS (Advanced Traveller Information Systems) are aimed at dynamically dispatching to travellers information about traffic conditions, allowing the travellers to take their travel decisions with less uncertainty. Moreover, even if the network effects of ATIS are still under discussion, it is generally argued that they could well lead to benefits also with reference to the transport system as a whole. In any case, the investigation of the information effects on travellers? behaviour is a pre-requisite for any analysis on the impacts of ATIS on transport networks.
One of the main aspects of travellers? behaviour that is influenced by ATIS is related to route choices. Several frameworks have been proposed in literature in order to incorporate travellers? reaction to information within route choice models. Some of these approaches have been based on Prospect Theory. In fact it has been shown, in several studies, that travellers exhibit some attitudes related to risk perception (risk aversion and risk seeking). Travellers? risk perception can be influenced by different factors. For instance, in many cases the reliability of travel times (i.e. variance of travel times across days) can be viewed as a source of risk. In networks where reliable and unreliable routes coexist, Prospect Theory can explain how some travellers exhibit a risk prone attitude and some others a risk averse one, which means that some travellers are inclined to choose shorter but less reliable routes while others longer but more reliable.
In addition to the travel time reliability, information accuracy (i.e. the ability of the system in predicting precisely traffic conditions) can also be considered in terms of risk perception: decreasing accuracy of the dispatched information can be perceived by travellers as increasing risk conditions. This phenomena has been studied in several studies, which characterised route alternatives by the different risk attitudes that are induced by travellers as a result of different information accuracies. Moreover, in the case of a differently accurate information context, travellers? behaviour is related to the relationship between the information accuracy and the degree of reliability of actual travel times.
The focus of this paper is the effect of ATIS accuracy on traveller's route-choice behaviour and risk perceptions in a simulated environment. The experiment has been carried on the basis of a Stated Preference (SP) methodology using an internet platform. A sample of 33 Italian respondents was recruited and invited to submit their answers via a web-site questionnaire (the final data base is composed of 990 records). Each respondent is repetitively required to make his/her travel choices in a simulated environment where an ATIS device (VMS - Variable Message Sign) dispatches travel information for several successive days.
During the course of the experiment three scenarios (in which independent groups of respondents are involved) are designed and tested. Respondents are provided with mixed information (descriptive and prescriptive information). Different information accuracy levels are tested by considering different levels of discrepancy between the ATIS-estimated travel times and the times the travellers will have actually experienced at the end of their trips. Travellers are also notified of their actual travel times at the end of each day allowing them to evaluate the accuracy of the ATIS ex-post, as well as progressively learn about the distributions over time and across alternatives of the actual travel times (i.e routes' reliability). In each scenario, the different levels of information accuracy can be perceived by respondent as a different condition of risk.
Preliminary analyses, based on non parametric statistical tests, revealed a systematic influence of different levels of information accuracy on the propensity of respondents to adopt risk-prone or risk-averse route-choice behaviour. This propensity is measured as the observed frequency with which respondents choose shorter (but with less reliable travel time) or longer (but more reliable) routes. These preliminary results show that a deeper disaggregated approach is worthwhile.
We adopt a discrete-choice approach to investigate based on the experimental data how accuracy of provided information affects route-choice behaviour. The choice made by each respondent can be regarded as compliant or non-compliant with the provided information and consequently as risk prone or risk averse. A mixed-logit model with panel data is applied in order to accommodate both the repeated-choice context of the experiment and the distribution of risk attitudes among respondents. In the paper the specification and estimation of the model framework will be fully presented and commented, as well as the performances of the approach in terms of sample reproduction. We note that this is a work in progress.
Association for European Transport