AI/ML Air Interface

Technical Explanation

3GPP Release 18 standardizes AI/ML integration into physical layer protocols. Neural networks running on the UE and gNodeB replace traditional algorithms: autoencoders compress CSI feedback matrices with 90% overhead reduction, while deep learning predicts beam blockage in mmWave.

Key Functions

• Standardized integration of deep learning models into 5G physical layers
• Employs autoencoder neural networks to compress CSI feedback by up to 90%
• Predicts dynamic beam blockage in millimeter wave channels in real time