Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLM-based Iterative Approach to Metamodeling in Automotive

Nenad Petrovic, Fengjunjie Pan, Vahid Zolfaghari, Alois Knoll

TL;DR

The paper tackles the challenge of keeping automotive metamodels aligned with evolving vehicle architectures, which is labor-intensive when done manually. It introduces an iterative LLM-based workflow that uses GPT-4o and Retrieval Augmented Generation (RAG) to generate Ecore metamodels and PlantUML visualizations from OEM requirements, with expert feedback guiding refinement. The prototype, implemented as a Python Flask web service, demonstrates end-to-end artifacts including partial and complete metamodels, updated parsers, XMI instances, and visualization pipelines, plus an optional OCL rule path. A comparative evaluation against a smaller local model shows GPT-4o yields syntactically correct Ecore results, while local models require additional transformations and feedback to reach similar usefulness. The work highlights potential time savings for automotive toolchains and outlines paths toward local deployment and future improvements in evaluation and model correctness.

Abstract

In this paper, we introduce an automated approach to domain-specific metamodel construction relying on Large Language Model (LLM). The main focus is adoption in automotive domain. As outcome, a prototype was implemented as web service using Python programming language, while OpenAI's GPT-4o was used as the underlying LLM. Based on the initial experiments, this approach successfully constructs Ecore metamodel based on set of automotive requirements and visualizes it making use of PlantUML notation, so human experts can provide feedback in order to refine the result. Finally, locally deployable solution is also considered, including the limitations and additional steps required.

LLM-based Iterative Approach to Metamodeling in Automotive

TL;DR

The paper tackles the challenge of keeping automotive metamodels aligned with evolving vehicle architectures, which is labor-intensive when done manually. It introduces an iterative LLM-based workflow that uses GPT-4o and Retrieval Augmented Generation (RAG) to generate Ecore metamodels and PlantUML visualizations from OEM requirements, with expert feedback guiding refinement. The prototype, implemented as a Python Flask web service, demonstrates end-to-end artifacts including partial and complete metamodels, updated parsers, XMI instances, and visualization pipelines, plus an optional OCL rule path. A comparative evaluation against a smaller local model shows GPT-4o yields syntactically correct Ecore results, while local models require additional transformations and feedback to reach similar usefulness. The work highlights potential time savings for automotive toolchains and outlines paths toward local deployment and future improvements in evaluation and model correctness.

Abstract

In this paper, we introduce an automated approach to domain-specific metamodel construction relying on Large Language Model (LLM). The main focus is adoption in automotive domain. As outcome, a prototype was implemented as web service using Python programming language, while OpenAI's GPT-4o was used as the underlying LLM. Based on the initial experiments, this approach successfully constructs Ecore metamodel based on set of automotive requirements and visualizes it making use of PlantUML notation, so human experts can provide feedback in order to refine the result. Finally, locally deployable solution is also considered, including the limitations and additional steps required.

Paper Structure

This paper contains 4 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Implementation Overview
  3. Experiment scenario and evaluation
  4. Conclusion

Figures (3)

  • Figure 1: LLM-based workflow for iterative metamodeling in automotive domain: 1-Freeform text 2-Text chunks 3-Ecore metamodel (partial) and updated model parser 4-Metamodel visualization based on PlantUML 5-Expert feedback/update request 6-Ecore metamodel (complete) 7-Freeform text 8-Model instance 9-XMI file 10-OCL rules based on standards 11-Complete and consistent system model 12-Executable code
  • Figure 2: Iterative metamodel construction example: 1) iteration 1-adding sensor requirements 2) iteration 2-adding actuator and power management requirements 3) iteration 3 - human expert feedback
  • Figure 3: Metamodeling workflow for locally deployable LLM