A UE measures reference signals, runs its part of a two-sided AI/ML model, estimates intermediate KPIs from the UE-side output, and reports them to a network node.

Patentradars sammanfattning

The invention defines a split AI/ML inference architecture for wireless networks in which a two-sided model is partitioned between a UE part and a network-node part. The UE measures reference signal resources specified by a dedicated reference signal configuration, performs local model inference, and derives intermediate Key Performance Indicators (KPIs) from its output. These intermediate KPIs are then reported to the network node, enabling the network side to complete joint inference or use the metrics for scheduling, resource management, or model monitoring without exposing raw model parameters.

AI-genererad och redaktionellt processad.