TITLE:
Comparison of Estimated Cycle Split Failures from High-Resolution Controller Event and Connected Vehicle Trajectory Data
AUTHORS:
Saumabha Gayen, Enrique D. Saldivar-Carranza, Darcy M. Bullock
KEYWORDS:
Split Failure, Connected Vehicle, Detector, Traffic Signal, Performance Measures, Trajectory Data
JOURNAL NAME:
Journal of Transportation Technologies,
Vol.13 No.4,
October
9,
2023
ABSTRACT: Current
traffic signal split failure (SF) estimations derived from high-resolution controller event data
rely on detector occupancy ratios and preset thresholds. The reliability of
these techniques depends on the selected thresholds, detector lengths, and
vehicle arrival patterns. Connected vehicle (CV) trajectory data can more
definitively show when a vehicle split fails by evaluating the number of stops
it experiences as it approaches an intersection, but it has limited market
penetration. This paper compares cycle-by-cycle SF estimations from both
high-resolution controller event data and CV trajectory data, and evaluates the
effect of data aggregation on SF agreement between the two techniques. Results
indicate that, in general, split failure events identified from CV data are
likely to also be captured from high-resolution data, but split failure events
identified from high-resolution data are less likely to be captured from CV
data. This is due to the CV market penetration rate (MPR) of ~5% being too low
to capture representative data for every controller cycle. However, data
aggregation can increase the ratio in which CV data captures split failure
events. For example, day-of-week data aggregation increased the percentage of
split failures identified with high-resolution data that were also captured
with CV data from 35% to 56%. It is recommended that aggregated CV data be used
to estimate SF as it provides conservative and actionable results without the
limitations of intersection and detector configuration. As the CV MPR
increases, the accuracy of CV-based SF estimation will also improve.