R1-2410733
discussion
FL summary #0 for AI/ML in beam management
From Samsung
Summary
This Samsung-moderated 3GPP RAN1 document (Tdoc R1-2410733) presents a comprehensive summary of AI/ML beam management contributions from meeting #118, containing over 100 proposals across multiple technical areas. The document covers UE-side and network-side models for both spatial (BM-Case1) and temporal (BM-Case2) beam prediction, addressing configuration, performance monitoring, inference reporting, and beam indication frameworks.
Position
Samsung advocates for a flexible framework combining multiple options for UE-side model applicability (merging Options 1 and 2), supports dedicated monitoring configurations separate from inference configurations, and pushes for practical beam accuracy indicators with multiple definition alternatives. They oppose overly restrictive single-option approaches and advocate for reusing existing CSI framework mechanisms where possible while introducing necessary AI/ML-specific enhancements.
Key proposals
- Proposal 1.1B (Sec 1.1): Potential working framework combining Options 1 and 2 for UE-side model applicability reporting, allowing both CSI-ReportConfig and inference parameter sets in Step 3
- Proposal 2.1-1a (Sec 2.1): Introduce beam accuracy indicator (BAI) for UE-assisted performance monitoring with multiple definition options (Top-1/1, 1/Top-K, Top-K/M, best of Top-K with margin)
- Proposal 2.2-1a (Sec 2.2): Support Option 2 for monitoring configuration using dedicated resource sets and report configuration in separate CSI report configuration linked to inference report
- Proposal 3.1a (Sec 3.1): For BM-Case2 reference time, support Option 3 based on latest transmission occasion of CSI-RS/SSB resource in Set B for measurement
- Proposal 3.4-1b (Sec 3.4): Configure two separate CSI-ResourceConfigIds for Set A and Set B when two resource sets are configured for UE-sided model
- Proposal 4.1b (Sec 4.1): Beam information in UE inference report uses CRI/SSBRI of Set A resource set, with ranking conveyed by order based on model output
- Proposal 6.1 (Sec 6): For NW-sided model BM-Case2, support reporting measurement results of multiple time instances in one report using differential L1-RSRP
- Proposal 6.2B (Sec 5.2): Support both options for NW-sided model training data collection - L1-RSRPs from resource sets and beam information of Top K beams
- Proposal 7.1 (Sec 7): Study extending Rel-17 TCI state activation/indication methods to activate N joint TCI states for N future time instances in BM-Case2
- Proposal 8.1 (Sec 8): Use existing CPU mechanism as starting point for UE-side AI/ML-based CSI processing with FFS on sharing between legacy and AI/ML reporting