Towards Safety-Compliant Transformer Architectures for Automotive Systems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kirchner, Sven, Purschke, Nils, Wu, Chengdong, Knoll, Alois
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912851357073408
author Kirchner, Sven
Purschke, Nils
Wu, Chengdong
Knoll, Alois
author_facet Kirchner, Sven
Purschke, Nils
Wu, Chengdong
Knoll, Alois
contents Transformer-based architectures have shown remarkable performance in vision and language tasks but pose unique challenges for safety-critical applications. This paper presents a conceptual framework for integrating Transformers into automotive systems from a safety perspective. We outline how multimodal Foundation Models can leverage sensor diversity and redundancy to improve fault tolerance and robustness. Our proposed architecture combines multiple independent modality-specific encoders that fuse their representations into a shared latent space, supporting fail-operational behavior if one modality degrades. We demonstrate how different input modalities could be fused in order to maintain consistent scene understanding. By structurally embedding redundancy and diversity at the representational level, this approach bridges the gap between modern deep learning and established functional safety practices, paving the way for certifiable AI systems in autonomous driving.
format Preprint
id arxiv_https___arxiv_org_abs_2601_18850
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Towards Safety-Compliant Transformer Architectures for Automotive Systems
Kirchner, Sven
Purschke, Nils
Wu, Chengdong
Knoll, Alois
Software Engineering
Transformer-based architectures have shown remarkable performance in vision and language tasks but pose unique challenges for safety-critical applications. This paper presents a conceptual framework for integrating Transformers into automotive systems from a safety perspective. We outline how multimodal Foundation Models can leverage sensor diversity and redundancy to improve fault tolerance and robustness. Our proposed architecture combines multiple independent modality-specific encoders that fuse their representations into a shared latent space, supporting fail-operational behavior if one modality degrades. We demonstrate how different input modalities could be fused in order to maintain consistent scene understanding. By structurally embedding redundancy and diversity at the representational level, this approach bridges the gap between modern deep learning and established functional safety practices, paving the way for certifiable AI systems in autonomous driving.
title Towards Safety-Compliant Transformer Architectures for Automotive Systems
topic Software Engineering
url https://arxiv.org/abs/2601.18850