Signal Awareness Applications

Mollenhauer, Mike; Viray, Reginald; Doerzaph, Zac; White, Elizabeth; Song, Miao · 2022 · ROSA P / Safety through Disruption (Safe-D) University Transportation Center (UTC)

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This research addresses the persistent safety issue of intersection collisions, which account for 40% of crashes on U.S. roadways, including approximately 800 annual fatalities from red-light violations. Despite existing infrastructure countermeasures, intersections remain high-risk zones, particularly for distracted drivers. The study aims to enhance the Virginia Connected Corridors (VCC), a connected vehicle test bed in Virginia, by developing signal awareness applications. Specifically, the project focuses on creating predictive Signal Phase and Timing (SPaT) models and a mobile application prototype to alert drivers to unsafe intersection approaches, such as red-light violation warnings. The methodology involved a five-step process to develop a prediction model using Long Short-Term Memory (LSTM) recurrent neural networks. Due to the adaptive nature of traffic controllers, which makes signal timing highly unpredictable, the team collected second-by-second historical data from the Smarterroads portal. This data, including controller settings, signal timing, and vehicle/pedestrian arrivals, was processed into 120-second sequences. The researchers trained LSTM models to predict the time remaining until signal state changes. They compared a model using only signal timing data against a comprehensive model incorporating all available data elements. Additionally, the team developed a mobile application graphical user interface (GUI) to display traffic light status, timing information, and vehicle location, utilizing lane-level accuracy supported by real-time kinematic position correction messages. The results demonstrated that the comprehensive LSTM model significantly outperformed the signal-timing-only model, reducing prediction error by a factor of four after 100 training epochs. Using a mean absolute error loss function, the final model achieved a prediction error of less than two seconds 83% of the time when trained on 39 days of data. The preliminary classification model, which predicted signal state at a fixed future point, was abandoned in favor of the regression model due to bias and limited utility. The mobile application prototype successfully displayed SPaT data and lane-specific countdowns when users were within 400 feet of an intersection. However, the live implementation of the predictive model faced performance constraints, preventing real-time integration into the existing server infrastructure during this phase. The significance of this work lies in its demonstration that machine learning can effectively predict adaptive signal timing with high accuracy, enabling more robust signal awareness applications. By integrating predictive SPaT data into a user-friendly mobile interface, the project provides a proof-of-concept for technologies that can curb unsafe driving behaviors, such as accelerating through yellow lights. The findings highlight the necessity of incorporating diverse data elements beyond basic timing to improve prediction accuracy. While real-time integration requires further optimization, the study establishes a foundation for future controlled evaluations of these safety features, potentially reducing intersection crashes and improving traffic flow efficiency.

Key finding

The comprehensive machine learning model incorporating all relevant data elements significantly outperformed a model using only signal timing data, reducing prediction error by a factor of four.

Methodology

modeling

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via bulk_ingest_rosap on 2026-05-23 (6 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• vru facing ehmi
• ehmi external hmi

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Methodological Resource: tool software