ISO 26262 ASIL-B Compliance for AI-Integrated HMI in Commercial Vehicle VCUs
Main Article Content
Abstract
The increasing deployment of artificial intelligence (AI) and machine learning (ML) within commercial vehicle human–machine interface (HMI) systems introduce functional safety challenges that conventional automotive safety standards do not yet fully resolve. Although ISO 26262 establishes a mature lifecycle framework for electrical and electronic systems, important questions remain regarding specification of learned behaviour, traceability between requirements and trained models, robustness under distribution shift, governance of training datasets, and assurance of probabilistic decision-making. These issues are particularly significant in heavy-duty commercial vehicles, where high vehicle mass, air-brake delay, trailer dynamics, and demanding duty cycles increase the operational consequences of delayed or misleading driver information. This paper proposes a practical ASIL-B compliance framework for AI-integrated HMI systems operating within commercial vehicle control units (VCUs). The framework combines ISO 26262 functional safety processes, ISO 21448 Safety of the Intended Functionality (SOTIF), and emerging AI governance guidance including ISO/IEC TR 5469, ISO/IEC 23894, ISO/IEC 42001, and the NIST AI Risk Management Framework. The proposed approach extends conventional automotive safety engineering through AI-aware hazard analysis and risk assessment (HARA), AI-extended design failure mode and effects analysis (DFMEA), fault tree analysis (FTA), structured verification and validation (V&V), and safety-case development tailored to bounded-authority AI-HMI architectures. The paper argues that ASIL-B is an appropriate integrity target when AI functionality remains advisory or supervisory and when deterministic safety mechanisms retain responsibility for critical warnings and fallback behaviour. The primary contribution is a systems-oriented ASIL-B compliance framework for AI-integrated HMI systems in heavy-duty commercial vehicle VCUs that incorporates AI failure modes, operational-domain assumptions, and bounded AI authority principles within an automotive functional safety context.