Enabling GLOSA through Domain Knowledge Aware SPAT Prediction and Queue Length Aware Trajectory Optimization

Project Abstract

This study extends efforts in improving Signal Phase and Timing (SPAT) prediction and Trajectory Optimization near signalized intersections. It focuses on efficiently modeling and addressing uncertainties in intersections controlled by actuated traffic signals. One major uncertainty is using deep learning for SPAT prediction. While deep learning performs well most of the time, there are instances of faulty predictions. To address this, the study combines deep learning with traffic signal domain expertise to ensure accurate predictions aligned with traffic signal controller logic. In trajectory optimization, uncertainties arise from predicting the waiting queue and its clearance time at traffic signals. This task involves complex factors like traffic conditions, vehicle dynamics, and driver behavior, including perception reaction time. Incorporating queue length estimation and clearance time into the trajectory planning algorithm will enable fuel-efficient optimization, particularly beneficial for Green Light Optimal Speed Advisory (GLOSA) during high traffic demands when queues have a significant impact. This study will conduct a comprehensive literature review to assess the current state of SPAT prediction and queue estimation, considering relevant publications in traffic signal control, machine learning, and optimization. The aim is to identify optimal approaches for incorporating domain knowledge into SPAT prediction and integrating queue estimation into trajectory planning.

Click here to read the full report. 

Universities Involved

Virginia Tech

Principle Investigators

Dr. Hesham Rakha

Amr Shafik

Seifeldeen Eteifa

Funding Sources and Amounts

USDOT: $100,000

Start Date

September 1, 2023

Completion Date

September 1, 2024

Expected Research Outcomes & Impacts

The study's final task involves preparing a comprehensive report documenting all steps and research findings. It will explain the research methodology, approaches used, and results obtained. The report will also offer recommendations for specific models and practices. The study investigates the potential of connected and automated vehicles optimizing trajectories to minimize fuel consumption. It explores the benefits and limitations of the GLOSA system application in connected and automated vehicles (CAVs). The developed algorithm can be integrated into CAVs' adaptive cruise control systems to promote eco-driving behavior. The final report will serve as a valuable reference for researchers and practitioners in the field of traffic optimization and fuel-efficient driving.

Subject Areas

Connected and Automated Vehicle, Infrastructure Planning and Design