Integration of Generative AI and Simulation Modeling: WSC 2026 Tutorial and Its Open-Source Practice Platform

In the intersection of artificial intelligence and system simulation, the deep integration of generative AI and traditional simulation modeling is opening up new possibilities. WSC2026 (Winter Simulation Conference) will launch a tutorial on this topic, and the supporting open-source project has been released—a Pyodide-based multi-stage interactive application that demonstrates the complete workflow from problem definition to agent AI demonstration, providing researchers and practitioners with a hands-on experimental platform.

Traditional simulation modeling is crucial in fields like manufacturing and logistics, but it faces challenges such as high modeling barriers, long development cycles, complex experimental design, and difficulty in result interpretation. Generative AI (e.g., large language models) can understand natural language requirements, automatically generate code, and explain results, making 'conversational simulation development' possible, lowering barriers, and improving efficiency.

The tutorial divides the simulation lifecycle into six phases:

1. Phase0: Problem Definition and Requirement Analysis (AI assists in translating requirements, identifying boundaries, and generating conceptual models);
2. Phase1a: Input Modeling and Data Preparation (AI analyzes data, generates missing value estimates, and cleans code);
3. Phase1b: Model Construction and Implementation (converting natural language to simulation code, generating components and documents);
4. Phase2: Model Execution and Operation (AI monitors status, dynamically adjusts parameters, and generates logs);
5. Phase3: Experimental Design and Result Analysis (AI automatically designs experiments, performs intelligent sampling, and explains results);
6. MCP Agent AI Demonstration (AI agents independently call simulations, multi-agent collaborative decision-making).

The supporting open-source project is a Pyodide-based multi-tab web application with the following features:

• Pyodide: In-browser Python environment, zero installation, supports scientific libraries, offline availability, secure sandbox;
• Multi-stage interactive interface: Modules divided by phases, automatic data transfer, support for saving and loading;
• Modular design: Easy to replace models, connect to different AI backends, and add custom components.

The platform has value in multiple scenarios:

• Education and training: Experiments without environment configuration, progressive learning, AI teaching assistants;
• Industrial consulting: Rapid proof of concept, interactive requirement clarification, automatic document generation;
• Academic research: Exploring new human-machine collaboration models, evaluating AI effects, building reproducible platforms;
• Enterprise internal use: Lowering modeling barriers, accelerating iteration, improving result interpretability.

When integrating generative AI into simulation, the following points need attention:

• Model accuracy and verification: Human-machine collaborative verification, automated testing, domain expert validation;
• Prompt engineering and context management: Custom prompt templates, managing dialogue context, optimizing prompts;
• Computing resources and response time: Optimizing Pyodide performance, asynchronous processing, timeout retry mechanisms.

The open-source project welcomes community contributions (bug reports, cases, AI integration improvements). Future outlook: More natural language-to-model conversion, intelligent experimental design, AI-driven insights, and widespread human-machine collaboration models.

The WSC2026 tutorial and open-source platform provide methodologies and tools for the integration of generative AI and simulation. Whether you are a senior practitioner, student, or technical personnel, it is worth trying. It demonstrates current capabilities and points to the future direction—the new era of AI-enhanced intelligent and efficient simulation is coming.