Edge Case Detection in Automated Driving: Methods, Challenges and Future Directions

arXiv cs.RO / 5/5/2026

💬 OpinionIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses a reliability gap in automated driving by surveying how to detect rare, unexpected “edge cases” that are difficult to cover in real-world data.
It provides a hierarchical classification of edge-case detection and assessment methods, first by AV modules (perception and trajectory-related subsystems covering prediction, planning, and control) and then by the underlying theories and methodologies.
It introduces and emphasizes “knowledge-driven” approaches that use expert insights and domain knowledge to find edge cases that may be missing from training datasets.
The survey also reviews evaluation techniques and metrics, spanning detection quality (precision/recall/false positives), real deployment constraints (compute overhead and detection delay), and domain-specific safety measures (crash rates and severity).
It highlights major remaining challenges—such as data quality/availability, validation and interpretability limits, the sim2real gap, and computational constraints—while outlining how the taxonomy can support modular testing and scenario generation.

Abstract

Automated vehicles promise to enhance transportation safety and efficiency. However, ensuring their reliability in real-world conditions remains challenging, particularly due to rare and unexpected situations known as edge cases. While numerous approaches exist for detecting edge cases, a comprehensive survey reviewing these techniques is lacking. This paper bridges this gap by presenting a hierarchical review and systematic classification of edge case detection and assessment methodologies. Our classification is structured on two levels: first, by AV modules, including perception and trajectory-related (encompassing prediction, planning, and control) sub-systems; and second, by underlying methodologies and theories guiding these techniques. Furthermore, we introduce "knowledge-driven" approaches, which complement data-driven methods by leveraging expert insights and domain knowledge to identify cases absent in training datasets. We then examine techniques and metrics for evaluating edge case detection methods, including detection performance (e.g., precision, recall, false positive rates), practical deployment (e.g., computational overhead, detection delay), and domain-specific measures (e.g., crash rates, severity analysis). We conclude by highlighting key challenges for edge case detection, including data availability and quality issues, validation and interpretability limitations, the sim2real gap, and computational constraints. The hierarchical classification and review of methods and assessment techniques in this survey enable modular and targeted testing frameworks by guiding the selection of detection methods for specific AV subsystems while considering methodological principles. It also supports practical testing by facilitating scenario generation in simulation and focused subsystem validation in the real world.