Human-LLM Interaction to Support MBSE Adoption in the Space Industry: Technical and Dialogic Perspectives

The usage of conversational interfaces powered by LLMs led to productivity gains in most industries and domains. General purpose solutions (e.g., ChatGPT, Claude) are easy to start with and offer decent speed-ups. But tailored solutions are gaining traction, promising to achieve more tasks with less user input, and to integrate into the Engineers’ working environment.

This session showcases applying LLMs (Large Language Models) to MBSE from two perspectives:

1. Technical:
SysDICE will present its work developing an assistant to help in key MBSE tasks, e.g., model understanding, modeling rules enforcement. A case study is built on top of a space-related reference SysML model using Cameo Systems Modeler.

2. Human–machine interaction:
Emilie Perreau (Grenoble University) will introduce the key interaction challenges (e.g., modeling human–LLM interaction components, specifying evaluation metrics, and analyzing implications for aerospace system engineering tasks) posed by generative conversational agents, before engaging in the co-refinement of a heuristic set to assess their impact on end users and inform design recommendations for Space systems engineering in MBSE activities.

Participants will also have the occasion to try an early-stage open-source application to support the selection of interactive evaluation criteria for LLM-based systems.

Schedule:

• Introduction and Scope (5 min)

• Overview of LLMs / LLM Workflows (10 min)

• Case Study: SysDICE AI4MBSE Assistant in Cameo Systems Modeler™ (20 min)

• Overview of Human–LLM Interaction Evaluation (10 min)

• Application and Co-design of Interactive Heuristics to Assess and Control the Ecological Impact of Generative Conversational Agents on Engineers’ MBSE Activities (20 min)

• Discussion, Q&A, and Reflection (25 min)