China Telecom Pilots AI Large Model for 5G Network Planning
China Telecom has successfully completed the first live-network pilot of a 5G wireless network planning large model. The pilot demonstrated a 50% increase in planning efficiency and achieved over 75% accuracy in automatically outputting design schemes. This shift from experience-dependent to model-driven planning automates complex infrastructure design and predicts construction outcomes. It accelerates the intelligent transformation of the telecommunications industry by replacing slow, manual site surveys with AI-driven digital twins. The system utilizes a dual-twin architecture combining a channel twin model for spatial reconstruction and a traffic twin model for predicting traffic fluctuations. In joint tests with ZTE in Shanghai, the model achieved over 80% accuracy for site planning in residential areas and over 75% accuracy in identifying coverage issues in underground parking lots.
## BACKGROUND
Traditional wireless network planning relies heavily on manual site surveys and expert analysis, which is time-consuming. Modern planning increasingly uses technologies like ray tracing to simulate radio wave propagation and Measurement Reports (MR) from user terminals to analyze network coverage.