R1-2407694
discussion
Discussion on AIML for beam management
From Spreadtrum
Summary
Spreadtrum presents their technical positions on AI/ML for beam management in NR, covering data collection, model inference, and performance monitoring aspects. The document contains 12 proposals and 3 observations addressing both UE-side and network-side AI/ML models for spatial and temporal beam prediction use cases.
Position
Spreadtrum advocates for UE-initiated data collection for UE-side models, opposes larger quantization steps for inference to maintain accuracy, supports up to 16 beams reporting per instance, and favors reusing existing CSI frameworks to minimize specification impact. They push against overly complex standardization solutions, advocating for simpler approaches like CRI/SSBRI-only beam information and implicit time reporting over more complex bitmap schemes.
Key proposals
- Proposal 1 (Sec 2.1.1): For UE-side model, support UE to request the data collection and report training-related information, such as expected measurement resources, etc.
- Proposal 2 (Sec 2.1.2): For the configuration of inference results reporting for UE-side model, not support only resource set for Set B is configured
- Proposal 3 (Sec 2.1.3.1): For NW-side model, for reported beam information, only support CRI/SSBRI of a measurement resource set
- Proposal 4 (Sec 2.1.3.2): For NW-sided model, both L1 and high-layer signaling can be used for data collection for training
- Proposal 5 (Sec 2.1.3.3): For BM-Case1 and BM-Case2 with a network-side AI/ML model, larger quantization step(s) should not be considered at least for model inference
- Proposal 7 (Sec 2.2.1): For BM-Case 1 and BM-Case 2, support UE to report the measurement results of up to 16 beams in one reporting instance
- Proposal 8 (Sec 2.2.1): Reporting multiple past time instances in one reporting instance for BM-Case2 is not needed
- Proposal 9 (Sec 2.2.2.1): For BM-Case2, implicit report of time information should be supported
- Proposal 10 (Sec 2.2.2.2): For BM-Case2 with a UE-side AI/ML model, the number of reported beams can adaptively selected by UE for each N instance
- Proposal 11 (Sec 2.3.1): support Alt.2 for Option 2 (UE-assisted performance monitoring)
- Proposal 12 (Sec 2.3.2): For inference for UE-side models, to ensure consistency between training and inference, option 1 should be considered
- Proposal 6 (Sec 2.2): For BM-Case2, TCI indication framework should be reused by gNB, e.g., beams from multiple time instance can be indicated to UE by multiple beam indications respectively