Generative AI-Powered Digital Twins for Crowd Management in Large-Scale Events

Managing safety and operational efficiency in large-scale events requires tools capable of capturing complex crowd dynamics while supporting rapid and informed decision-making. This paper presents a Generative AI-powered digital twin framework that integrates agent-based crowd simulation, an API-based execution pipeline, and a Large Language Model (LLM)-driven conversational interface within a unified system. The proposed approach enables dynamic configuration, execution, and analysis of crowd scenarios under varying operational conditions, including high-demand and emergency evacuation contexts. Experimental results demonstrate the system’s ability to reproduce nonlinear crowd dynamics, detect congestion patterns, and evaluate evacuation performance, providing actionable insights for planning and safety assessment. A key contribution lies in the introduction of an API-based execution paradigm that exposes the full simulation lifecycle (configuration, validation, execution, and output retrieval) through programmatic interfaces, enabling reproducible and scalable what-if analysis. Additionally, the integration of an LLM-based conversational interface allows non-technical users to interact with complex simulation models through natural language, significantly improving accessibility and usability. The framework is validated through a TRL-4 prototype, demonstrating robust performance, scalability, and interaction reliability. Overall, the proposed system advances digital twins from static analytical tools to executable, interactive, and user-centric platforms for decision support in complex urban environments.