UL 4600: What to Include in an Autonomous Vehicle Safety Case

作者：P. Koopman · 发表于：Computer · 年份：2023 · DOI：10.1109/mc.2023.3236171 · 被引用次数：27 · 研究领域：Computer Science

Autonomous vehicles (AVs) will not see widespread use until we can be sure that they are acceptably safe. That remains a big challenge, but robust support from safety standards can help. To illustrate what is involved in ensuring AV safety, this article provides an overview of the approach taken by the ANSI/UL 4600 AV safety standard.1 MITIGATING UNUSUAL HAZARDS While AV safety involves software, hardware, sensor technology, and more, it is software safety that is by far the biggest challenge. Creating life-critical software goes well beyond writing high-quality code. Even perfect software implementations might have safety issues due to requirements defects and encounters with novel operational environments. Adding the complexity of machine learning-based technology brings into play issues such as avoiding training data bias and ensuring a sufficiency of data samples for rare events. While progress on AV functionality to date has been impressive, it might be said that deploying the first hundred vehicles is the easy part. Ensuring the lifecycle safety of a truly driverless AV, deployed at scale across a wide variety of operational conditions, will require significant engineering effort that has only just begun. Conventional vehicles place a significant safety burden on the driver for handling unusual circumstances Digital Object Identifier 10.1109/MC.2023.3236171 Date of current version: 3 May 2023 UL 4600: What to Include in an Autonomous Vehicle Safety Case