SimGPT: Unlocking potentials of generative AI in simulation via Simio and LLM integration

Generative Artificial Intelligence (GenAI) is transforming decision support systems, particularly in simulation modeling, by enabling automation, optimization, and analytics. This paper introduces SimGPT, a framework leveraging GenAI across generation, execution, and analysis stages in Discrete Event Simulation. SimGPT integrates the Simio simulation environment with Large Language Models (ChatGPT, Gemini), enabling real time AI-powered decision support within simulation models. Case studies demonstrate the framework’s versatility across: (1) real time job dispatching where LLM based decisions outperformed traditional heuristics, (2) prompt engineering showing critical design factors for effective LLM integration, (3) comparative analysis of LLM models and versions, and (4) synthetic data generation for emergency room simulations including patient arrivals, characteristics, and length of stay prediction. These findings showcase GenAI’s potential to enhance simulation methodologies. This work pioneers GenAI’s application in simulation, opening new research opportunities. • Introduces SimGPT, a framework integrating GenAI with simulation environments. • Develops an API connecting LLMs like ChatGPT and Gemini to simulation software. • Demonstrates real-time adaptation, data generation, and predictive modeling. • Advances automation, scalability, and reasoning in simulation-based decision support. • Pioneers GenAI applications in simulation, opening new avenues for research.