Wiz Technology, a Shanghai-based company, launched its Fidget spatiotemporal foundation model in Zhangjiang, Shanghai, along with an urban agent system called Spatiotemporal Metaverse for city forecasting and automated operations.

The company said Fidget was trained on more than 20,000 categories of multimodal data collected from Zhangjiang AI Innovation Town, including geographic information, enterprise activity, pedestrian flow, transportation, electricity consumption, events, work orders, and camera feeds.

In tests in the town, Wiz said the model achieved 90% accuracy in pedestrian-flow prediction, 95% in transportation prediction, and 97% in parking prediction.

At the launch event, the system used pedestrian-flow and weather forecasts to send a drone 1.5 kilometers to a road, where it identified accumulated garbage, returned images for verification, generated a cleanup work order, and dispatched an unmanned cleaning vehicle.

Wiz said the project was developed under the guidance of the Shanghai Pudong New Area Data Bureau.

Researchers generally describe spatiotemporal foundation models as AI systems trained on heterogeneous space-time data that can be adapted to multiple forecasting and planning tasks, rather than a single application.

Chinese researchers are also exploring city-scale agent platforms such as CityGPT.

🔗 Source: Security Times