This repository is part of the research conducted for the paper:
Advanced User Interaction with Urban Digital Twins using Large Language Models
Authors: Khaoula Kanna, Thomas H. Kolbe
Link: will be available after publication.
Published at: GSW 2025 Conference
This repository demonstrates the implementation of a framework integrating Large Language Models (LLMs) with Urban Digital Twins (UDTs). The framework supports intuitive interaction with semantic 3D City Models (CityGML) and time-series data (SensorThings API). The implementation highlights:
- Querying geospatial and dynamic data via natural language.
- Mapping user queries into SQL for CityGML data in 3DCityDB.
- Visualization of results.
The repository is structured to show case researchers, city planners, and citizens the potential of using LLMs for urban decision-making and understanding.
A list of example queries can be found in queries.md.
- CityGML: The CityGML LOD2 model was stored in the 3DCityDB. The 3DCityDB Schema given to the LLM can be found here: database_schema.json
- SensorThings API: SensorThings API GitHub Repository
This work is supported by the ASCEND EU-project, in collaboration with the City of Munich.
For further inquiries, feel free to reach out to:
- Khaoula Kanna: [email protected]
- Prof. Dr. Thomas H. Kolbe: [email protected]