Conference Papers 2022

Seminar
Day 1 (7 Sep 2022), Session 2, UNDERSTANDING MULTIMODAL TRAVEL PATTERNS VIA DATA, 14:00 - 16:00

Status
Accepted, documents submitted

Submitted by / Abstract owner
Chris Jacobs-Crisioni

Authors
Chris Jacobs-Crisioni, European Commission, Joint Research Centre
Eric Koomen, VU University Amsterdam
Yosef König, VU University Amsterdam

Short abstract
Floating car data, describing speeds on a very high spatial and temporal resolution, were used to estimate models of door-to-door travel time unreliability. To do so, trajectories were created from those data for every minute on 120 selected paths in

Abstract
This paper presents a study in which so-called floating car data were used to estimate door-to-door travel time unreliability in the Netherlands. This study extends prior investigations into travel time unreliability that were based on speed detection loops on motorways and camera tracking on secondary roads (Kouwenhoven, Mulder, & Van Mourik, 2015). It was executed with the aim of refining unreliability models used for in societal cost benefit analyses using the Netherlands national road traffic model. For this purpose, the perceived advantages of using floating car data are that they offer a more comprehensive overview of network speeds at a finer spatial and temporal resolution, so that they enable door-to-door travel time estimates and offer more insight in travel speeds in particular on the secondary road network.
The used floating car data have been obtained predominantly from the users of a mobile phone application. This application essentially trades up-to-date route information on congestion and speed controls for information on users’ movement and travelled speeds. These data describe, for every minute of the day and for every roughly 50 metres long segment on the road network it is linked to, the average speed from the 10 most recent travel speeds on any road segment; at least if many speed observations are available on the segment. The data does come with some caveats. Observed speeds are curated, so they are never presented as higher than the legal maximum speeds on a segment. If insufficient observations are available within the minute, additional observations are drawn from observed speeds in the prior 30 minutes.
To convert the available data into door-to-door travel time estimates, trajectories were recreated using time and segment-specific travel speeds. These trajectories were created for every minute between 6 AM and 8 PM in a four month period, for a set of 120 paths. Paths were selected through an elaborate procedure that picks paths that are representative for trips in the Netherlands, have minimal overlap, and maximal coverage in the floating car data. They connect locations that roughly coincide with the locations where the national traffic model’s zones connect to its road network. Following the mentioned previous study into unreliability, travel time estimates for every minute for every path were subsequently aggregated to single observations for every quarter hour, representing the expected travel time for that quarter hour given a well-informed traveller; expected delay; and regular unreliability. These data were subsequently used to estimate models of how unreliability responds to expected delays.
This presentation will go deeper into the data generation process and the results from this study. The opportunities offered by data such as used here, as well as some of its limitations, will be used as well.
Reference: Kouwenhoven, M., Mulder, M., & Van Mourik, H. (2015). Betrouwbaarheid van reistijden in het LMS/NRM. Tijdschrift Vervoerswetenschap, 51(1), 3–16.

Programme committee
Data