arXiv cs.AI· Johann Maximilian Christensen, Elena Hoemann, Frank K\"oster, Sven Hallerbach·· 5 小时前AI 评分32
基于数据为安全关键 AI 系统定义运行设计域（ODD）的新方法
Defining Operational Conditions for Safety-Critical AI-Based Systems from Data
AI 导读
一项 arXiv 研究提出用多维核表示从已有数据中事后定义安全关键 AI 系统的运行设计域（ODD），以解决传统方法依赖专家知识导致的 ODD 描述不完整问题。该方法可全自动推导、置换稳定、构造上有界，并在仿射等变归一化下对单位选择不变；论文还给出了两个 ODD 等价的相似性定义并证明了体积收敛性。方法经合成基准与真实航空用例验证，可支持数据驱动安全关键 AI 系统的未来认证。
正文 · 原文
Abstract:Artificial Intelligence (AI) has been on the rise in many domains, including numerous safety-critical applications. However, for complex systems in the real world, defining the underlying environmental conditions in which the AI-based system must operate---the Operational Design Domain (ODD)---is extremely challenging. This often results in an incomplete description of the ODD, which contrasts with the requirements of many domains for certifying AI-based systems. Traditionally, the ODD is created in the early stages of the development process, drawing on sophisticated expert knowledge and related standards. This paper presents a novel method for defining the ODD a posteriori from previously collected data using a multidimensional kernel-based representation. This approach is validated through both synthetic benchmarks and a real-world aviation use case. Moreover, the paper defines similarity of two ODDs if they generate the same outputs up to Lebesgue-null input sets and proves convergence in volume of the calibrated representation under the stated assumptions. The novel, safety-by-design, deterministic kernel-based ODD representation is derived fully automatically, given documented assurance inputs, permutation-stable, bounded by construction, and, under affine-equivariant per-dimension normalization, invariant to the choice of units. Utilizing the proposed ODD representation supports future certification of data-driven, safety-critical AI-based systems.
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2601.22118 [cs.AI] |
| (or arXiv:2601.22118v4 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2601.22118 arXiv-issued DOI via DataCite |
Submission history
From: Johann Christensen [view email]
[v1]
Thu, 29 Jan 2026 18:46:02 UTC (161 KB)
[v2]
Wed, 6 May 2026 12:23:08 UTC (140 KB)
[v3]
Mon, 31 Aug 2026 08:25:06 UTC (227 KB)
[v4]
Wed, 28 Oct 2026 09:35:21 UTC (373 KB)
来源：arXiv cs.AI · arxiv.org