NMC Tackles Map-Matching to Create App-Ready Transit Files
The Network Modeling Center (NMC) has developed innovative, open-source software (Map Matcher) to match up maps that were not designed to be used together, enabling use of those maps in regional transit planning. NMC specializes in traffic modeling that helps regions forecast traffic patterns to ease congestion.
NMC researchers Kenneth A. Perrine, Natalia Ruiz Juri, and Alireza Khani were faced with the problem of adapting General Transit Feed Specification (GTFS) files into the NMC’s existing traffic models. These files are provided by many transit agencies to publicly give out transit schedule and route information in a way that software packages and web-based services, such as Waze or Roadify, can use them. As part of a CAMPO initiative to bring advanced modeling to the Central Texas region, the NMC wanted to use the transit functionality of FAST-TrIPs (a relatively new transit assignment model developed by Dr. Khani and his PhD advisor at the University of Arizona) within its traffic models, but faced the problem of map compatibility. In particular, it was challenging to integrate Khani’s FAST-TrIPs work (which depends upon GTFS schedule information and bus routes) with the NMC’s dynamic traffic assignment (DTA) models.
The core of the problem is that the DTA model represents a region’s roadway network by a graph of links attached among geographically placed nodes. Generally, links are where vehicles travel, and nodes are intersections or end-to-end joints (shown on the right in the figure below). Meanwhile, GTFS files provide only a series of GPS coordinates that describe a route (shown on the right in the figure below). These coordinates, however, cannot easily be mapped to the underlying node-link-style network because of a number of problems that often crop up: error in the precision of GPS coordinates or the underlying map; missing georeference points; missing underlying topology features such as bridges, tunnels, and tight corners. The topic of map-matching has been around for a while and tackled by quite a number of researchers with varying degrees of success, but none had addressed the problem in the exact way that our researchers did, nor provided their algorithm to the planning community as open-source software.
At the 94th Annual TRB Meeting in January 2015, Perrine presented the Map Matcher software at a transit planning-related poster session. “I started talking to people about fifteen minutes before the poster session began, and was talking to a constant stream of people until all of my neighbors had taken down their posters after the session was ended,” Perrine noted as he recounted his discovery of how many researchers within various areas of transportation wrestle with very similar problems. “Everyone was excited about the idea of using open-source software to solve their problems.”
While Perrine will agree that the software is not perfect and still a work in progress, he welcomes opportunities for collaborations in making improvements. The software repository and documentation is available on GitHub through http://ctr.utexas.edu/nmc/nmc-map-matcher.