Overview
This project explores the use of commercially available consumer GPS data in travel reliability studies. Travel time reliability concerned in this study has two dimensions. The first dimension is the probability of completing a trip within a given time budget, the so-called on-time arrival probability. This measure is related to how the decision maker deems the importance of a trip. The second dimension has to do with the fact that for the trip deemed as equally important (i.e. the same on-time arrival probability is required), two individuals may choose different routes and reserve different amounts of time for travel depending on their risk-taking preference. This dimension reflects traveler’s risk-taking behavior in the face of uncertainty, which is largely ignored in previous studies.
This project examines two TomTom traffic data products, namely MultiNet and TrafficStats, which produce respectively time-of-day link travel speed profile and travel time statistics on selected routes. The project has two primary objectives. The first is to evaluate the quality of the aforementioned commercial data products using publically available traffic sensor data (specifically, Gary-Chicago-Milwaukee, or GCM, database). The second objective has to do with evaluating the impacts of this new data source on reliable route guidance. The findings from our case study are:
- TomTom speed profile data generate average travel time estimations for highway segments that match those from GCM data reasonably well. However, estimated average link travel times from TomTom are about 10% – 15 % lower than those from GCM data.
- Using TomTom data as a benchmark, we found that the travel times on arterial streets are severely underestimated in the GCM database. Note that these arterial travel times were estimated because no direct observations were available.
- The daily speed profiles obtained from TomTom and GCM data do not match very well on most expressways examined in this study. In general, TomTom data tend to underestimate the travel speeds on expressway segments. It appears that, in most cases, TomTom caps the speed at the legal limit, which is not the true free flow speed, as revealed from the GCM data.
- The quality and usefulness of the data provided by TrafficStats raise more concerns. For one thing, the percentile route travel times provided in the route reports seems to spread out too much and have variances too high to be realistic in most cases. More problematic is the fact that these percentile travel times do not match the reported route segment statistics, especially in terms of variances. The conclusion of the research team is that one has to consider the utility of this product with cautions, especially given its high price.
- The reliable routing experiments conducted in this project show that the reliability routing results are changed significantly after TomTom data are used to generate travel time distributions on the arterial streets. The main reason for this change is that the travel times on arterial streets in GCM were severely underestimated. As a result, many arterial-based paths that were found reliable for certain risk-taking preferences were no longer attractive because they become much longer when TomTom data are used. This finding suggests that MutliNet speed profile may be used to as a supplementary data source for travel reliability studies.
The project generates a new TomTom database that stores all TomTom data acquired in this project, as well as a GCM database. Both databases are managed using PostgreSQL system on a server hosted at Northwestern University. The project also develops an application, called RIDEA (or Reliability-IDEA) based on the VNET platform. VNET is a simple, flexible and extensible graphic user interface that supports a wide variety of network-related applications. RIDEA is available for download, along with the VNET platform, at http://translab.civil.northwestern.edu/nutrend/?page_id=53. RIDEA provides a graphical interface to access and visualize the databases and to conduct reliable routing experiments. It may be used as a prototype to support future commercial software development for travel reliability analysis.