A Machine Learning Perspective on Automated Driving Corner Cases
A Machine Learning Perspective on Automated Driving Corner Cases
🔍 Preprint, 2025
Overview
Autonomous driving must be safe — but how do you define a “corner case”? Traditional approaches enumerate difficult scenarios one by one, an approach that doesn’t scale and ignores the data distribution that ML models actually learn from.
We propose a fundamentally new distributional perspective on corner cases. Instead of manually cataloging edge cases, our framework reasons about them through the lens of the underlying data distribution — making corner case recognition automatic, principled, and scalable.
Key Contributions
🔀 Unifies existing taxonomies — subsumes scenario-based corner case categorizations under a single distributional framework
📊 Strong benchmark performance — achieves competitive results on corner case detection across standard OOD benchmarks (extended for this purpose)
🌫️ Combined corner cases — introduces a new fog-augmented Lost & Found dataset for analyzing compound corner cases
🚫 No manual specification needed — corner cases are identified from the data distribution, not hand-crafted rules
Why It Matters
Safe autonomous driving requires detecting the unexpected. This work replaces brittle, hand-crafted corner case lists with a principled ML framework — a scalable foundation for building truly robust perception systems.
For high-stakes applications, like autonomous driving, a safe operation is necessary to prevent harm, accidents, and failures. Traditionally, difficult scenarios have been categorized into corner cases and addressed individually. However, this example-based categorization is not scalable and lacks a data coverage perspective, neglecting the generalization to training data of machine learning models. In our work, we propose a novel machine learning approach that takes the underlying data distribution into account. Based on our novel perspective, we present a framework for effective corner case recognition for perception on individual samples. In our evaluation, we show that our approach (i) unifies existing scenario-based corner case taxonomies under a distributional perspective, (ii) achieves strong performance on corner case detection tasks across standard benchmarks for which we extend established out-of-distribution detection benchmarks, and (iii) enables analysis of combined corner cases via a newly introduced fog-augmented Lost & Found dataset. These results provide a principled basis for corner case recognition, underlining our manual specification-free definition.
@article{Schmidt2025d,title={A Machine Learning Perspective on Automated Driving Corner Cases},author={Schmidt, Sebastian and K\"{o}rner, Julius and G\"{u}nnemann, Stephan},year={2025},month=oct,journal={ArXiv},volume={2510.10653},url={http://arxiv.org/abs/2510.10653},}