AI/ML in the air interface and network management.

8 technologies in AI-RAN and Open RAN.

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.