Bridging the Gaps: A Comparative Analysis of ISO 21434, ISO 26262 and Machine Learning in Autonomous Vehicles

• Suelen Daiene de Oliveira Bastos USP
• Kalinka Castelo Branco USP
• André Luiz de Oliveira UFJF

Resumo

The integration of Machine Learning (ML) in Au-tonomous Vehicles (AV s) raises challenges not fully addressed by current automotive standards, specifically ISO 26262 (functional safety) and ISO/SAE 21434 (cybersecurity). This paper reviews gaps in these standards concerning ML-enabled systems, focusing on explainability, data lifecycle, runtime assurance, and ethical compliance. To bridge these gaps, we propose the ML-Aware Compliance Bridge, a four-layer framework covering: (i) Data Lifecycle Management, (ii) Model Governance, (iii) Runtime Assurance, and (iv) Ethics and Explainability. The framework maps standard deficiencies to actionable recommendations, sup-porting the development of safer and more trustworthy AI -driven autonomous systems.

Palavras-chave: Bridges, Ethics, Runtime, Reviews, ISO Standards, Machine learning, Safety, Standards, Robots, Autonomous vehicles, ISO/SAE 21434, ISO 26262, Machine Learning