Open and Extendable Urban Digital Twins: Real-Time Data Pipelines and Conversational Interfaces for Sustainable Cities

Cities face increasing sustainability pressures, yet many smart-city tools remain fragmented or limited to single domains. Urban Digital Twins (UDTs) offer an integrated alternative by combining real-time data, visualization, and analytics, but most implementations still lack scalability, accessibility, and cost efficiency. This paper presents an open and modular UDT prototype for the city of ‘s-Hertogenbosch. The system integrates a Kafka-based ingestion pipeline, InfluxDB for time-series storage, and a CesiumJS 3D city model. To improve accessibility, we developed ODIN, a conversational interface that connects to a hybrid local–cloud large language model and enables natural-language queries. The prototype overlays simulated CO₂ and noise streams with live traffic, weather, and air-quality feeds and forecasts. In a construction-logistics scenario, the streaming pipeline processed simulated sensor events with latencies of 6–14 ms and query response times of 55–150 ms. ODIN’s first-pass translation accuracy on these simulated datasets ranged from 30% with local models to 85–100% with hosted GPT/DeepSeek backends. The prototype demonstrates how open-source architectures can support affordable municipal UDT deployments, with future work focusing on IoT integration, predictive analytics, and city-scale stress testing.