Real-time Prediction of Queue at Signalized Intersections to Support Eco-driving Applications
Principal Investigator: Dr. Mecit Cetin
Abstract: Predicting the anticipated traffic conditions in the downstream is a critical component in developing eco-driving applications to smooth the acceleration and braking behavior of vehicles. This research aims to develop new models to predict downstream conditions in a connected-vehicles environment where equipped-vehicles exchange their locations, speed, and other relevant information. In particular, new models will be developed to predict the queue dynamics at signalized intersections based on both the real-time information supplied by equipped vehicles and the signal phase and timing (SPAT) provided by controllers. Based on the predicted or anticipated downstream conditions, equipped vehicles can then regulate their speeds to adjust their trajectories to minimize fuel consumptions and emissions. The developed prediction models will be tested in a microsimulation environment (e.g., VISSIM) under various scenarios including under-saturated and oversaturated conditions, different market penetration levels for equipped vehicles, different vehicle classes, etc. In addition, the accuracy of the models will be tested as the prediction horizon is varied (e.g., from zero to several cycle lengths). To the extent possible, the models will also be tested on field data collected as part of the NGSIM effort. The results will help in identifying traffic conditions and market penetration levels under which queue lengths can be predicted effectively to support eco-driving strategies.