Real-Time Turning Movement, Queue Length, and Traffic Density Estimation and Prediction Using Vehicle Trajectory and...

Area of Science:

• Intelligent Transportation Systems
• Traffic Engineering
• Control Theory

Background:

• Real-time traffic signal control requires accurate estimation and prediction of traffic states.
• Existing methods often struggle with data limitations and varying market penetration levels of connected vehicles.

Purpose of the Study:

• To introduce a novel two-stage adaptive Kalman filter algorithm for enhanced traffic state estimation and prediction.
• To improve the accuracy and reliability of turning movement counts, queue sizes, and traffic density estimations.
• To validate the algorithm's performance using real-world and simulated data.

Main Methods:

• A two-stage adaptive Kalman filter algorithm was developed.
• Stage 1: Estimates turning movement (TM) counts using probe vehicle trajectory and upstream detector data.
• Stage 2: Estimates upstream approach density and queue sizes.
• Evaluation involved drone-collected and simulated data from a signalized intersection.

Main Results:

• The Kalman filter significantly improved traffic state estimation accuracy compared to baseline methods.
• Stage 1 TM estimates showed up to a 50% improvement in standard deviation.
• Stage 2 queue size estimation improved by up to 32.8%, and traffic density by up to 18.5%.
• Reliable predictions were achieved even at low market penetration levels (e.g., 5%).

Conclusions:

• The proposed two-stage adaptive Kalman filter effectively enhances traffic state estimation and prediction for real-time signal control.
• The algorithm demonstrates robustness across various market penetration levels, indicating practical applicability.
• This approach offers a significant advancement for intelligent transportation systems and traffic management.