AI/ML in the air interface and network management.

Technologies

• AI/ML for CSI feedback Learned compression and prediction of channel state information to improve MIMO precoding.
• AI/ML beam management Predicting best beams in space and time to cut measurement overhead, specified from Rel-19.
• AI/ML positioning Direct and assisted learned positioning that improves accuracy in non-line-of-sight conditions.
• AI/ML for NG-RAN functions Learned load balancing, mobility optimisation and energy saving between RAN nodes.
• AI/ML model lifecycle management Training, testing, deployment, monitoring and retirement of models inside the managed network.
• Network data analytics Core analytics function producing predictions and statistics consumed by RAN and core functions.
• O-RAN AI/ML workflow Where models are trained, hosted and inferred across SMO, RICs and network functions.
• 5G system support for AI/ML applications Network features for model transfer, split inference and federated learning traffic.