Virtual Detection at Intersections using Connected Vehicle Trajectory Data

Abstract

Performance measures and adaptive control methodologies for traffic signal systems currently require intersections to be instrumented with vehicle detectors and communication equipment, which can require substantial engineering resources to deploy and maintain. Recent studies have explored the use of Connected Vehicle (CV) data for signal performance measures at various levels of market penetration, with some results showing possibilities for analyzing coordinated operations with penetration rates as low as 1–5%. The current study explores whether CV trajectory data (based on the crowdsourcing of timestamped vehicle and mobile device locations) can currently be used for this purpose. The study compares in-pavement detector data with “virtual” detector line data based on geo-fencing of trajectory splines. The Purdue Coordination Diagram (PCD), arrival flow profile, and cumulative arrival performance measures are used to assess and compare both data sources. The comparisons indicate that percent on green (POG) estimates from CV trajectories come within 0.1 to 15.7% of the in-pavement detector data, with the highest accuracy attributing to time-of-day periods with moderate traffic and high POG. Market penetration rate of CV in this study vary between approximately 0.6 to 2%.