Estimating on-road vehicle density using crowdsourced data and Monte Carlo analysis

摘要

Mobile sources are a major source of urban air quality pollution. Modeling mobile source emissions is challenging due to the varying nature of vehicle densities at any given time period for any given road segment. The standard method of evaluating this uncertainty is to perform resource intensive traffic counts and apply the results from one investigation to similar roadways. Crowdsourced traffic information has been collected and used by Google Maps since 2009, providing near real-time traffic conditions in the form of colored line segments superimposed over roads. Using this colored traffic status on Google Maps, we present a novel method of estimating traffic densities on road segments that account for different times of use and road types. Using screenshots centered on specific segments, we determine the traffic condition based on local pixel color values and link the segment to a road type (freeway/ highway or surface road) with associated maximum speeds (120 kph for highways and 80 kph for surface roads). After evaluating the traffic conditions of the segment based on its color status, the number and types of vehicles on the segment are estimated using a Monte Carlo technique based on estimated traffic speed, vehicle registration records, and local driving patterns. Once traffic densities are quantified, emissions from the segments can be estimated and modeled. Sampled segments can be aggregated to evaluate time-series densities of entire networks in cities and regions.