Automated Generation of Simulation Models for Production and Logistics Processes Using LLM-Based Multi-Agent Systems

Discrete-event simulation (DES) is a well-established method for analyzing and optimizing complex production and logistics systems. However, its use is often limited by high modeling effort and the need for specialized expertise. This article presents a novel multi-agent system based on Large Language Models (LLMs) that automates the creation and validation of simulation models. Building on a previous approach of ours, the new system employs an agent-based architecture designed to address issues such as context loss and the need for manual validation. Specialized agents handle tasks from requirements elicitation to results evaluation. The implementation uses open source frameworks "LangGraph" to structure agent interactions and "SimPy" to model the simulation logic. A case study demonstrates that the system can automatically and reproducibly model a complete production scenario from text-based descriptions. The results show realistic modeling and a significantly reduced modeling effort compared to both a manual approach and our previous system. The proposed approach lowers reliance on expert knowledge and makes simulation-based methods more accessible to non-specialist users.