The ability to passively track large numbers of mobile devices has generated a lot of excitement in recent years. Traditional travel surveys without GPS tracking, either wholly or in part, seems passe today. However, GPS tracking can only overcome some of the problems associated with travel diary surveys. It can capture missed trips and route choice information, but at increased cost per survey. However, it does nothing to ease the rising cost and difficulty associated with contacting, recruiting, retaining households in the first place, or collecting and processing the data. Passive tracking overcomes these limitations, at the expense of giving up interaction with the device owner. Thus, information about the traveler, trip purpose, and other details must be inferred or lost.

The ideal solution is to use both travel diaries and passive tracking together. While techniques for fusing these data are yet to be proven the concept has strong intuitive appeal. Until such techniques emerge the question becomes whether such data can be used on their own, and whether they resemble the output of travel models built using traditional survey data. The opportunity to answer such questions arose recently in the Research Triangle Region of North Carolina, which served as a pilot project for using origin-destination data from AirSage as an adjunct to modeling. A detailed analysis of the differences between the AirSage data and the Triangle Regional Model (TRM) has been completed, as well as a comparison of both data sources to travel survey data. The differences were subtle in some cases and surprising in others. Several approaches for using the AirSage data were tested.

This presentation will report on the findings of this comparison. A brief description of the AirSage data and TRM will provide the context for the discussion. The methodology for comparing the two data sources will be presented, as will the results. Methods used to train and constrain the TRM using the AirSage data will be also be discussed. Recommendations for the use of passive tracking data in modeling practice will conclude the presentation.