Re-imagining ISO 26262 in the Age of Autonomous Vehicles: Enhancing Controllability through Transferability and Predictability

A new research paper proposes enhancements to ISO 26262 functional safety standards to address autonomous vehicles operating at SAE Levels 4-5, where human drivers are absent. The framework introduces Transferability and Predictability as measurable sub-concepts to replace the traditional Controllability metric, enabling falsifiable safety claims across different operational design domains.

The ISO 26262 standard has long served as the foundation for automotive functional safety, but its risk assessment framework assumes human drivers can intervene in failure scenarios. As autonomous vehicles advance toward full autonomy without human operators, this assumption becomes invalid, creating a critical gap between legacy standards and modern safety requirements. This research addresses that disconnect by decomposing Controllability into two auditable dimensions that reflect AV operational realities.

Transferability captures how effectively autonomous systems can hand control to designated fallback safety mechanisms, while Predictability measures how reliably external agents—passengers, pedestrians, other vehicles—can anticipate AV behavior. By grounding Predictability in human-robot interaction principles and providing mathematical frameworks for quantification, the authors make safety claims verifiable rather than abstract. The introduction of a designed-versus-achievable gap distinguishes theoretical architectural fallback capabilities from real-world, scene-dependent performance.

For autonomous vehicle developers and regulators, this framework offers practical advantages. It aligns with both ISO 26262 and ISO/PAS 21448 (SOTIF) standards, enabling consistent safety documentation across different operational design domains. Manufacturers can use these metrics to substantiate safety claims through traceable, falsifiable evidence rather than relying on subjective assessments. This approach reduces regulatory uncertainty while accelerating certification processes for Level 4-5 systems.

The framework preserves existing standards rather than replacing them, making adoption less disruptive. As regulators worldwide struggle to establish safety requirements for driverless vehicles, this research provides a methodological bridge between traditional automotive safety practices and the technical realities of fully autonomous systems. Early adoption by major manufacturers could establish industry precedent.

• →ISO 26262 requires modification to address autonomous vehicles without human drivers by replacing Controllability with Transferability and Predictability metrics.
• →The proposed framework makes AV safety claims falsifiable and traceable across different operational design domains rather than subjective.
• →Transferability measures fallback mechanism activation while Predictability quantifies external agents' ability to anticipate AV behavior using mathematical frameworks.
• →The research preserves existing standards while extending applicability to SAE Levels 4-5 autonomous systems, reducing regulatory uncertainty.
• →A designed-versus-achievable gap distinguishes theoretical safety capabilities from actual scene-dependent performance in real-world AV operations.