R1-2410367
discussion
Discussions on AI/ML for beam management
From CAICT
Summary
This CAICT document proposes 10 technical proposals for AI/ML beam management in 5G NR, covering UE-sided model performance monitoring, reference signal configuration for Set A/B, and measurement reporting for network-sided models.
Position
CAICT advocates FOR comprehensive UE-sided model monitoring with flexible network configuration (supporting both periodic and event-triggered reporting), per-cell based Associated ID design, and separate CSI-ResourceConfigId configuration (Alt 3) for Set A/B. They push FOR finer quantization steps in L1-RSRP reporting (less than 3dB accuracy) and flexible UE implementation of monitoring algorithms, positioning against overly rigid standardization of UE-side monitoring procedures.
Key proposals
- Proposal 1 (Sec 2.1): For option 2 of type 1 UE-sided model monitoring, the metric of UE reporting should include the accuracy of RSRP estimation, accuracy of predicted top 1/K beam and other information if necessary
- Proposal 2 (Sec 2.1): A time window could be configured by network to UE for UE-sided model monitoring and the detail algorithms for model monitoring at UE side could be left for UE implementation
- Proposal 4 (Sec 2.1): Both periodical or event triggered monitoring report could be considered and periodical monitoring report should be considered as baseline
- Proposal 5 (Sec 2.1): Event-1 could be used for the triggered event for Option 1 and/or Option 2
- Proposal 6 (Sec 2.2): Associated ID should include Set A/B configuration and per-cell based design could be considered as starting point. Other configurations could be FFS
- Proposal 7 (Sec 2.2): For configuration of Set A and Set B for UE-sided model at least for BM Case-1, Alt 3 are proposed for further discussion
- Proposal 8 (Sec 2.4): UE should follow the configuration from NW to report measurement results of Set B and/or Set A for NW sided model training data collection
- Proposal 9 (Sec 2.4): For subset determination of NW-sided model report, Alt.1 (Top M beam(s) is the beams with largest M measured values of L1-RSRPs) is proposed
- Proposal 10 (Sec 2.4): Finer step size should be considered for differential L1-RSRP feedback for NW-sided model report
- Proposal 3 (Sec 2.1): For option 2 of type 1 UE-sided model monitoring, the configuration of Set A/B for inference could be reused as default and whether dedicate configuration for Set A/B for monitoring is needed or not could be FFS