Variability of traffic-flow measures across freeway lanes

Abstract

On most freeways, a number of factors interact to produce lane-to-lane variations in speed and volume which are both site and volume dependent. The following paper explores and statistically verifies these variations using detector data and a combination of complementary techniques based on data collected for 30 days at 27 detector stations in May 1994 along the Queen Elizabeth Way freeway near Toronto, Ontario. The analysis indicates that considerable volume dependencies exist at each site, and that these dependencies are site specific. In addition to their independent variations, speed and flow are also shown to interact differently across different lanes and result in different underlying speed-flow relationships. The findings are intended to be relevant to the calibration of microscopic traffic simulation models and automatic incident detection algorithms. As such, the paper does not attempt to specifically identify the underlying causes for the variations, but rather attempts to recognize the aggregated effects of these causes in a fashion that would be useful to real-time freeway traffic management system control strategies relying solely on loop detector inputs.Key words: traffic-flow theory, traffic modeling, traffic simulation, incident-detection algorithms.