Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Description of Corner Cases in Automated Driving: Goals and Challenges

Daniel Bogdoll, Jasmin Breitenstein, Florian Heidecker, Maarten Bieshaar, Bernhard Sick, Tim Fingscheidt, J. Marius Zöllner

TL;DR

This paper addresses the challenge of corner cases (CC) in automated driving and argues for machine-interpretable corner-case descriptions (CCD) to improve offline data analysis and online system validation. It surveys scenario description languages (e.g., PEGASUS layers, OpenScenario/OpenDrive, ontologies) and CC-related tasks (detection, generation, dataset engineering), and outlines six research questions to guide CCD development. The authors illustrate CCD potential with three CC use cases—glare, unknown objects, and occluded objects—demonstrating how CCD can inform sensor fusion, coverage analysis, and synthetic data generation. The work outlines a roadmap for integrating CCD into perception, prediction, and planning, enabling more robust offline data curation and online deployment for automated driving.

Abstract

Scaling the distribution of automated vehicles requires handling various unexpected and possibly dangerous situations, termed corner cases (CC). Since many modules of automated driving systems are based on machine learning (ML), CC are an essential part of the data for their development. However, there is only a limited amount of CC data in large-scale data collections, which makes them challenging in the context of ML. With a better understanding of CC, offline applications, e.g., dataset analysis, and online methods, e.g., improved performance of automated driving systems, can be improved. While there are knowledge-based descriptions and taxonomies for CC, there is little research on machine-interpretable descriptions. In this extended abstract, we will give a brief overview of the challenges and goals of such a description.

Description of Corner Cases in Automated Driving: Goals and Challenges

TL;DR

This paper addresses the challenge of corner cases (CC) in automated driving and argues for machine-interpretable corner-case descriptions (CCD) to improve offline data analysis and online system validation. It surveys scenario description languages (e.g., PEGASUS layers, OpenScenario/OpenDrive, ontologies) and CC-related tasks (detection, generation, dataset engineering), and outlines six research questions to guide CCD development. The authors illustrate CCD potential with three CC use cases—glare, unknown objects, and occluded objects—demonstrating how CCD can inform sensor fusion, coverage analysis, and synthetic data generation. The work outlines a roadmap for integrating CCD into perception, prediction, and planning, enabling more robust offline data curation and online deployment for automated driving.

Abstract

Scaling the distribution of automated vehicles requires handling various unexpected and possibly dangerous situations, termed corner cases (CC). Since many modules of automated driving systems are based on machine learning (ML), CC are an essential part of the data for their development. However, there is only a limited amount of CC data in large-scale data collections, which makes them challenging in the context of ML. With a better understanding of CC, offline applications, e.g., dataset analysis, and online methods, e.g., improved performance of automated driving systems, can be improved. While there are knowledge-based descriptions and taxonomies for CC, there is little research on machine-interpretable descriptions. In this extended abstract, we will give a brief overview of the challenges and goals of such a description.

Paper Structure

This paper contains 16 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Related Work
  3. Scenario Description
  4. PEGASUS Layer Model
  5. Open Scenario & Open Drive
  6. Operational Design Domain
  7. Corner Case Detection and Generation
  8. Corner Case Detection:
  9. Corner Case Generation:
  10. Dataset Engineering
  11. Research Questions
  12. Addressed Use Cases
  13. Glare Effect
  14. Unknown Objects
  15. Occluded Objects
  16. ...and 1 more sections

Figures (3)

  • Figure 1: ML pipeline in automated driving with an emphasis on corner cases and their relation to data and ML models based on corner case descriptions.
  • Figure 2: A visually challenging scene due to glare Heidecker2021b (l) and an unknown object in the field of view chan2021segmentmeifyoucan (r).
  • Figure 3: Sequence of temporarily occluded road users Cordts2016.