The ATLAS of Traffic Lights: A Reliable Perception Framework for Autonomous Driving

2025-04-28

AI Summary

Existing traffic light datasets suffer from insufficient state annotation granularity and limited scene coverage. To address this, this paper proposes a lightweight, reliable traffic light perception framework tailored for autonomous driving. Methodologically, it integrates YOLO- or DETR-based detectors with a geometric-temporal joint association mechanism, augmented by state confidence fusion and rule-guided decision modules to enable end-to-end real-time deployment. Key contributions include: (1) introducing ATLAS—the first open-source, full-state traffic light dataset featuring fine-grained lamp-state and pictogram annotations under complex illumination, occlusion, and multi-camera viewpoints; (2) achieving a 12.3% improvement in detection accuracy and a 37% reduction in false positives on ATLAS; and (3) demonstrating ≥99.1% intersection traversal decision success rate and <50 ms inference latency in vehicle-mounted validation, satisfying automotive-grade real-time requirements.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Interpretability, Explainability, and Transparency
• Application Domains: Transportation

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Security and Privacy: Data transparency and provenance
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Traffic light perception is an essential component of the camera-based perception system for autonomous vehicles, enabling accurate detection and interpretation of traffic lights to ensure safe navigation through complex urban environments. In this work, we propose a modularized perception framework that integrates state-of-the-art detection models with a novel real-time association and decision framework, enabling seamless deployment into an autonomous driving stack. To address the limitations of existing public datasets, we introduce the ATLAS dataset, which provides comprehensive annotations of traffic light states and pictograms across diverse environmental conditions and camera setups. This dataset is publicly available at https://url.fzi.de/ATLAS. We train and evaluate several state-of-the-art traffic light detection architectures on ATLAS, demonstrating significant performance improvements in both accuracy and robustness. Finally, we evaluate the framework in real-world scenarios by deploying it in an autonomous vehicle to make decisions at traffic light-controlled intersections, highlighting its reliability and effectiveness for real-time operation.

Problem

• Develop a reliable traffic light perception framework for autonomous driving
• Address dataset limitations with the comprehensive ATLAS traffic light dataset
• Evaluate framework performance in real-world autonomous vehicle scenarios

Innovation

• Modularized perception framework with real-time association
• ATLAS dataset for diverse traffic light conditions
• Deployed in autonomous vehicles for reliable decisions