On failures of RGB cameras and their effects in autonomous driving applications
DOI: 10.1109/issre5003.2020.00011
archive: archived pipeline: cataloged verified
Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)
Summary
This paper investigates the failure modes of RGB cameras in autonomous driving systems and their subsequent impact on vehicle safety. While RGB cameras are critical sensors for tasks like object detection and lane recognition, their susceptibility to degradation poses significant risks. The authors address the lack of systematic analysis regarding realistic camera failures by defining specific failure modes, reviewing existing mitigations, and quantifying the performance drop in AI-based driving agents when processing corrupted images. The study employs a Failure Modes and Effects Analysis (FMEA) on the five primary components of an RGB camera: the lens, camera body, Bayer filter, image sensor, and Image Signal Processor (ISP). This analysis identified a comprehensive set of failure modes, including environmental factors (rain, ice, condensation), mechanical issues (broken lens, blur), and electronic faults (dead pixels, ISP failure). To evaluate the impact of these failures, the authors developed a Python library to simulate these defects on images from the KITTI Vision Benchmark Suite. They tested six different object detectors on these altered images to measure detection capability via Average Precision. Additionally, they injected these failures into a self-driving agent within a simulated environment to observe behavioral missteps. The results demonstrate that camera failures significantly degrade the performance of object detectors and cause unsafe behaviors in autonomous driving simulations. The study provides quantitative evidence that certain failures, such as those affecting image clarity or color accuracy, lead to substantial reductions in detection precision. The authors also highlight that while mitigations exist for individual failures, there is currently no orchestrated approach to handle the entire set of potential camera malfunctions. The developed software library allows for the reproduction of 130 different failure configurations, providing a resource for robustness assessment. The significance of this work lies in its contribution to the design of safe and robust autonomous vehicle architectures. By providing a systematic classification of camera failures and empirical data on their effects, the paper offers a methodology for evaluating the robustness of image-based AI/ML applications. The findings underscore the necessity of considering camera reliability in safety-critical domains and provide reference data to help developers identify the most dangerous failure modes. This supports the development of countermeasures and improves the overall safety of autonomous driving systems reliant on visual inputs.
Provenance
The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.
| Stage | Outcome | Tool | Model | Prompt | Attempts | Completed |
|---|---|---|---|---|---|---|
| discover | success | Crossref | — | — | 1 | 2026-06-17 |
| archive | success | unpaywall | — | — | 2 | 2026-06-25 |
| extract | success | cached | — | — | 2 | 2026-06-26 |
| clean | success | clean | — | — | 1 | 2026-06-20 |
| chunk | success | chunk | — | — | 1 | 2026-06-20 |
| embed | success | embed | Qwen/Qwen3-Embedding-8B | — | 1 | 2026-06-20 |
| enrich | success | openalex | — | — | 1 | 2026-06-20 |
| promote | success | — | — | — | 1 | 2026-06-17 |
| summarize | success | llm | qwen3.6-27b-prismaquant | summ-v5 | 1 | 2026-06-26 |
| tag | success | vector_similarity | — | — | 6 | 2026-06-20 |
| verify | success | — | — | — | 1 | 2026-06-26 |
Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.
Topics
Ranked by relevance to this paper. Hover a topic for its definition.