Real-Time Prediction of Queues at Signalized Intersections to Support Eco-Driving Applications. Final Report.

摘要

The overall objective of this research is to develop models for predicting queue lengths at signalized intersections based on the data from probe vehicles. The time and space coordinates of the probe vehicles going through signalized intersections are utilized to predict the back of the queue profile. For a single intersection, prediction models are developed where both over-saturated and under-saturated conditions are considered. The shockwave theory (i.e., the Lighthill-Whitham-Richards theory) is used to estimate the evolution of the back of the queue over time and space from the event data generated when probe vehicles join the back of the queue. An analytical formulation is developed for determining the critical points required to create the time-space diagrams that characterize queue dynamics. These critical points are used to estimate the queue lengths. The formulation is tested on the data obtained from traffic simulation software VISSIM. It was found that the shockwave-based formulation is effective in estimating queue dynamics at signalized intersections for--and over-saturated conditions even with a relatively low percentage of probes (e.g., 10-20%) in the system. For example, under over-saturated conditions simulated, the error is less than -10% in more 90% of the cycles when the market penetration of probe vehicles is 15%.