Khalifa University Launches RF-GPT to Interpret Wireless Signals
Khalifa University of Science and Technology's Digital Future Institute has released RF-GPT, an AI language model trained to analyze radio-frequency signals directly. The model converts wireless signals into visual patterns that AI systems can process and describe in plain language, addressing a gap in telecom AI where existing models work only with text and structured network data.
RF-GPT outperformed baseline models by up to 75.4% on radio frequency spectrogram tasks. It identified the number of signals in a spectrogram roughly 98% of the time-a task general-purpose AI models rarely accomplish.
How It Works
The model was trained on approximately 625,000 computer-generated radio signal examples. It handles tasks including signal type identification, detection of overlapping transmissions, recognition of wireless standards, Wi-Fi device usage estimation, and extraction of data from 5G signals.
By making the physical layer queryable in natural language, the model enables AI systems to support network optimization and policy decisions without requiring specialized RF expertise.
Research and Applications
Professor Merouane Debbah, Senior Director of the Digital Future Institute, led the project alongside post-doctoral fellows Hang Zou and Yu Tian, research scientists Dr. Lina Bariah and Dr. Samson Lasaulce, and researchers from Zhejiang University.
The work is designed for telecom operators, network engineering teams, and spectrum authorities managing increasingly complex wireless environments. It contributes to the UAE National Artificial Intelligence Strategy and provides a foundation for AI-native 6G networks.
Professor Ahmed Al Durrah, Associate Provost for Research at Khalifa University, said the launch reflects the institution's focus on digital infrastructure innovation aligned with national priorities.
Learn more about Generative AI and LLM applications and research in specialized AI domains.
Your membership also unlocks:
