What is it about?
This study examines the safety of monocular depth estimation systems used in self-driving cars, focusing on how these systems perform at urban intersections. By analyzing a typical four-way intersection scenario in daylight, we identify potential failures—like distance estimation errors or missed detections—and evaluate their impact. The research also proposes ways to mitigate these risks and sets safety thresholds to improve system reliability.
Featured Image
Photo by Brock Wegner on Unsplash
Why is it important?
Self-driving cars rely heavily on accurate depth perception to navigate safely, especially in busy urban areas. Our work is the first to systematically analyze failure risks for monocular depth estimation at intersections, providing actionable insights for engineers and policymakers. By quantifying these risks and offering mitigation strategies, we help advance the development of safer autonomous vehicles and contribute to future safety standards.
Perspectives
Constructing a structured approach to uncover unseen failure modes—just by observing vehicle behavior was a profound learning experience. It revealed how deeply interconnected and nuanced the relationships are between hardware-level perception sensors and software-level algorithms in mitigating risks. This project underscored that safety in autonomous systems isn’t just about individual components but about their seamless collaboration. I hope this work sparks further dialogue on bridging gaps between theoretical risk analysis and real-world deployment, ultimately making self-driving technology more transparent and trustworthy for the public.
Dr. Sanjay Singh
Manipal Institute of Technology, Manipal
Read the Original
This page is a summary of: FMEA-Based Safety Analysis of Monocular Depth Estimation for Autonomous Vehicles, February 2025, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/aide64228.2025.10987392.
You can read the full text:
Contributors
The following have contributed to this page
