Mapping low-resolution edges to high-resolution paths: the case of traffic measurements in cities

TL;DR

The paper tackles the problem of mapping low-resolution traffic measurements onto a high-resolution urban street network by linking each measured edge to a shortest-path sequence in the street graph. It introduces a multi-criteria matching framework that considers endpoints’ proximity, path length, angular variation, and geometric proximity (path surface), using $k$-nearest-node filtering and generating $2k^2$ candidate paths per measurement edge. Applied to Paris data with $k=4$, the method achieves a high matching rate (~99%), while providing quantitative scores and visual diagnostics to detect data discrepancies and limit-case failures, with human verification advised for complex interchanges. The study demonstrates a practical, extensible approach for leveraging open street data in traffic-network analysis, while identifying data integrity and temporal-consistency as key challenges for real-world deployment.

Abstract

We consider the following problem : we have a high-resolution street network of a given city, and low-resolution measurements of traffic within this city. We want to associate to each measurement the set of streets corresponding to the observed traffic. To do so, we take benefit of specific properties of these data to match measured links to links in the street network. We propose several success criteria for the obtained matching. They show that the matching algorithm generally performs very well, and they give complementary ways to detect data discrepancies that makes any matching highly dubious.

Figures (9)

• Figure 1: Overlayed drawings of OSM Paris street network (in black) and traffic sensor network (in blue).
• Figure 2: Cumulative distribution of link length (in meters) for both networks in Paris
• Figure 3: Normalized scores for each criterion with all 5591 matched edges ranked in ascending score order (logarithmic scale of the y-axis is used on the right)
• Figure 4: Correlation scores (normalized to [0,1] for both axis each time) of paths matched using a criterion (on the x-axis) with all other criteria one by one (on the y-axis) : first line is LC, then RC, SC, AC. Red dots correspond to correlation with RC, green with SC, blue with AC, orange with LC. Scores are normalized : both axis are [0,1]. Hence the first red plot on top left corner is the correlation between LC and RC scores for edges matched by minimizing LC. The green one just on its right is the correlation between LC and SC scores from matching minimizing LC, and so forth. To understand how to analyze this figure, from the correlation between LC and RC (top-left corner) we focus on the few reds dots around (1,0). These dots are edges with good RC score, but extremely bad LC score (though LC was the focus of the matching here) : this path has very little angular variation yet it is way too short or long.
• Figure 5: Output of the matching for RC, where the lowest-rated edges for each criteria are highlighted (50 for LC/AC, 300 for RC/SC due to scores on Figure \ref{['fig:scores']}).