Stateful Multi-Agent LLMs for Cross-View Interface Alignment in Automotive Model-Based Systems Engineering

2026-08-08

AI Summary

This work addresses the challenge of architectural drift in automotive model-based systems engineering caused by probabilistic inconsistencies between behavioral and structural views when using large language models. To ensure cross-view interface alignment, the authors propose a state-preserving multi-agent validation pipeline that innovatively integrates adversarial auditing with state-backtracking mechanisms. Semantic consistency is enforced through a class→activity→sequence ordered generation matrix grounded in the Vehicle Signal Specification (VSS) via retrieval-augmented generation (RAG). An independent AI validation agent dynamically audits outputs and triggers a state-preserving backtracking loop to resolve incompatibilities. Evaluated on an ADAS scenario, the approach achieves 97% entity traceability, 87% signal conservation, and an 85% F1 score—substantially outperforming standard RAG, which yields 0% traceability—and represents the first method to achieve deterministic alignment across heterogeneous architectural views.

Abstract

While Large Language Models (LLMs) can accelerate Model-Based Systems Engineering (MBSE) for software-defined vehicles, their probabilistic nature causes "architectural drift", fabricating interfaces in behavioral views that lack structural foundations. To enforce deterministic interface alignment, we propose a stateful, multi-agent validation pipeline. The framework utilizes a sequential generation matrix (Class->Activity->Sequence) and Vehicle Signal Specification (VSS)-grounded Retrieval-Augmented Generation (RAG). An independent AI Validator Agent dynamically audits outputs against a strict error taxonomy, triggering state-preserving backtracking loops to resolve incompatibilities. Evaluated on an Advanced Driver Assistance System (ADAS) scenario, standard RAG yielded 0% Entity Traceability. Conversely, our multi-agent workflow eradicated cross-phase hallucinations, achieving 97% Entity Traceability, 87% Signal Conservation, and an 85% F1-score. This proves adversarial auditing enables LLMs to reliably synthesize zero-error MBSE architectures.

Problem

• architectural drift
• interface alignment
• Model-Based Systems Engineering
• Large Language Models
• cross-view consistency

Innovation

• stateful multi-agent
• cross-view alignment
• retrieval-augmented generation
• architectural drift mitigation
• MBSE validation