R1-2410376
discussion
Discussion on AI/ML for beam management
From NTT DOCOMO
Summary
NTT DOCOMO's technical document presents 20 comprehensive proposals for AI/ML-based beam management in 5G NR, covering both UE-sided and network-sided model aspects including configuration, inference, performance monitoring, and data collection optimization.
Position
DOCOMO advocates for a pragmatic, network-centric approach that prioritizes operational efficiency and overhead reduction. They strongly push FOR: reusing existing CSI frameworks, network-controlled decision making (type 1 monitoring only), statistical performance reporting, and payload optimization techniques. They push AGAINST: type 2 performance monitoring, supporting multiple alternatives for the same function, and ignoring network operational burden in dynamic applicability scenarios.
Key proposals
- Proposal 1 (Sec 2.1): Resources of Set A and Set B should be configured in CSI-ResourceConfig, with Set A for training/monitoring and Set B for inference/monitoring
- Proposal 2 (Sec 2.2): Associated ID should be configured within one resource set for Set A/B to assume similar Tx beam properties for each RS within the resource set
- Proposal 4 (Sec 2.3): Support configuration under CSI-ReportConfig with one CSI-ResourceConfigId for Set B and another for Set A
- Proposal 5 (Sec 2.3): Support either always CRI/SSBRI or bitmap/CRI/SSBRI hybrid approach based on overhead optimization, but not both alternatives
- Proposal 9 (Sec 2.3): Support payload overhead reduction using absolute RSRP for best combination and differential RSRP for remaining combinations
- Proposal 11 (Sec 2.4): Not support type 2 performance monitoring, only type 1 where gNB makes decisions based on UE reporting
- Proposal 13 (Sec 2.4): Support Alt 3 performance metrics to check accuracy of predicted RSRP values in actual field conditions
- Proposal 16 (Sec 2.4): Support reporting of statistical performance metric values over multiple samples in UE-assisted performance monitoring
- Proposal 18 (Sec 2.5): Consider network operation burden due to dynamic applicability updates when discussing activation/deactivation restriction rules
- Proposal 19 (Sec 3.1): Consider overhead reduction for >4 beam information using large quantization step size and multi-time-instance reporting
- Proposal 20 (Sec 3.2): Introduce enhancements when UE-side conditions like Rx beam assumptions make network-side beam prediction difficult