ESTIMATION OF QUEUE LENGTH AND GREENLIGHT DURATION AT SIGNALIZED INTERSECTION USING COLLABORATIVE PERCEPTION AND MACHINE LEARNING METHODS IN V2X ENVIRONMENTS

Document Type

Conference Proceeding

Publication Date

1-1-2025

Abstract

Connected and Autonomous Vehicles (CAVs) utilize ecodriving systems for efficient fuel consumption while traversing signalized intersections. Applications such as Green Light Optimal Speed Advisory(GLOSA), Eco Arrival and Departure, and other Eco- Driving support these objectives. Estimating the accurate signal phase timing durations with real-time evaluation of queue lengths, queue clearance timings, and other traffic flow parameters play a crucial role in enabling energy-efficient maneuvering through intersections. To enable the dynamic processing of traffic flow, this research focuses on utilizing collaborative perception through Vehicle-to-Everything (V2X) communication to estimate queue lengths and broadcast estimated green light durations through enhanced SPAT messages. In our approach, Road Side Units collaborate with CAVs through perception data sharing to calculate green light duration at signalized intersections, enabling efficient traffic flow through optimized broadcasting of timing information. The study features a random forest algorithm with geo-spatial and corridor enhancements. Our model is trained on corridor specific data from Metro Detroit area, capturing individual intersection geometry and signal timing plans. Results reveal our system efficiently determines duration of green light based on vehicle distribution along the lanes leading to the intersection, queue length and traffic flow parameters. This efficiency is maintained under different traffic volumes and queue lengths with improved performance.

Publication Title

Proceedings of the ASME Design Engineering Technical Conference

ISBN

[9780791889213]

Link to Full Text

Share

COinS