NR Group Beam Management With Spatial Filter Prediction

The patent applies preliminary action by performing spatial filter prediction in advance using AI/ML models. The network device predicts spatial filters for future time instances based on historical data and current channel conditions, preparing beam configurations before they are actually needed. This allows the system to maintain optimal beam management performance while reducing real-time computational complexity and downlink resource overhead, as predictions are made proactively rather than reactively.

The patent implements self-service by enabling the terminal device to perform spatial filter prediction locally using AI/ML models. Instead of relying on the network device to provide all prediction computations, the terminal device autonomously predicts spatial filters using its own processing capabilities. This distributes the computational burden, reducing network downlink resource overhead and overall system power consumption while maintaining improved beam management performance.

The patent applies merging by combining multiple beam predictions into a unified group beam (pair) prediction framework. Instead of independently managing individual beam predictions, the system groups related spatial filters and predicts them collectively using AI/ML models. This consolidation improves beam management efficiency by handling multiple beams simultaneously while reducing overall system complexity through unified processing and resource allocation.