R1-2410359
discussion
Discussions on specification support for beam management
From Sharp
Summary
Sharp's contribution discusses comprehensive specification support for AI/ML beam management in NR Release 19, covering both UE-side and network-side models for spatial (BM-Case 1) and temporal (BM-Case 2) beam prediction. The document contains 23 technical proposals and 1 observation across performance monitoring, resource configuration, measurement reporting, inference results, and beam indication.
Position
Sharp advocates for comprehensive AI/ML beam management support using existing CSI frameworks as baseline while proposing specific enhancements: they strongly support Alternative 3 (two separate CSI-ResourceConfigIds for Set A and Set B) over other alternatives, favor beam prediction accuracy as the performance metric, and push for flexible configuration options including aperiodic CSI-RS support and probability information reporting. They oppose artificial restrictions on computational resource separation and advocate for shared CPU usage between legacy and AI/ML processing.
Key proposals
- Proposal 1 (Sec 2.1.1): Define beam prediction accuracy metric as ratio of accurately predicted beams where if predicted Top 1/K beams match actual Top 1 or Top-K measured beams, they count toward accuracy
- Proposal 4 (Sec 2.2): Support two CSI-ResourceConfigIds configured for Set A and Set B separately in CSI report configuration for UE-side model
- Proposal 8 (Sec 2.3.1): Support configuring more than 4 measurement results (e.g., L1-RSRP) in one CSI report as UE capability for NW-side model BM-Case 1
- Proposal 12 (Sec 2.4): Support beam information on predicted Top K beams and probability related information with gNB-configurable quantization step size for UE-side model inference results
- Proposal 18 (Sec 2.5): Reuse CSI framework for model performance monitoring for UE-side model
- Proposal 21 (Sec 2.6): Support shared overall CPU between legacy CSI reporting and AI/ML-based CSI reporting for UE-side model
- Proposal 23 (Sec 2.7): Enhance unified TCI state framework to support beam indication of multiple future time instances for BM-Case 2
- Proposal 11 (Sec 2.3.2): Support reporting measurement results from multiple past time instances in single CSI report configuration with differential L1-RSRP reporting for NW-side model BM-Case 2
- Proposal 13 (Sec 2.4): Support single L1-RSRP corresponding to largest predicted value across multiple time instances with remaining as differential RSRPs
- Proposal 2 (Sec 2.1.2): Additional contents beyond measurement results unnecessary for option 1 of Type 1 performance monitoring for UE-side AI/ML model
- Proposal 6 (Sec 2.2): Support Aperiodic CSI-RS for CMR at least for Set B inference in BM Case-2
- Proposal 22 (Sec 2.6): Consider setting higher priority for L1-RSRP reports in Rel-19 AI/ML BM
- Proposal 15 (Sec 2.4): Reference time based on latest transmission occasion of CSI-RS/SSB resource in Set B for measurement with configurable offset
- Proposal 19 (Sec 2.5): CSI report configuration can indicate purpose as inference or monitoring for UE-side model