FMEA-Based Safety Analysis of Monocular Depth Estimation for Autonomous Vehicles

Deploying monocular depth estimation systems in autonomous vehicles presents unique challenges, particularly in complex urban environments. This study conducts a comprehensive Failure Mode and Effects Analysis (FMEA) focusing on autonomous vehicle behavior at urban intersections. We identify and categorize critical failure modes, their effects, and potential mitigation strategies by analyzing a single four-way intersection scenario under daylight conditions. The research systematically quantifies risks using established metrics, including depth accuracy and processing latency. Our findings reveal that the most critical failure modes occur during the intersection entry phase, with distance estimation errors and missed detections presenting the highest Risk Priority Numbers (RPNs). We establish safety thresholds and validation frameworks for intersection scenarios through detailed performance metrics and statistical analysis. This focused approach provides valuable insights for system designers and safety engineers, contributing to developing robust safety guidelines for monocular depth estimation systems.