A Novel Map-Matching Algorithm for Relating Work Zones and Crashes

Modeling traffic crashes within work zones is critical to crash analysis and prevention, and it requires spatially relating each crash to a specific work zone if applicable. Currently, this is achieved by finding the corresponding work zone recorded for each crash in the crash database, or by assigning a crash to the work zone with the closest reference point to it. The challenges are twofold: (1) crash data can be limited and incomplete. Only a few large work zones are recorded in the crash database; and (2) reference points of a work zone do not accurately represent the work zone area, especially in cases of long work zones along curved roadways. Both issues would cause either inaccurate or infeasible mapping of crashes to work zones, which was overlooked in the work zone related crash analysis in the past. In this paper, a novel map-matching algorithm is developed to accurately and automatically map crashes to their respective work zone projects utilizing their combined attributes, including geographic coordinates, traffic flow directions, and route numbers. The algorithm is tested using the Pennsylvania Crash database and Pennsylvania Roadway Conditions Reporting System (RCRS) data from 2015 to 2017. We show that the new algorithm can substantially reduce the localization error of crashes mapped to work zones from hundreds of meters to meters comparing to existing algorithms. In addition, the proposed algorithm is computationally efficient, which takes tens of hours for mapping all crashes in the state-level roadway networks on a regular personal computer.