R1-2410737
discussion
FL summary #4 for AI/ML in beam management
From Samsung
Summary
This is Samsung's summary document (R1-2410737) for AI/ML in beam management from RAN1 #119 meeting, containing over 200 proposals and observations from multiple companies covering UE-side and NW-side model configurations, performance monitoring, data collection, and beam indication frameworks.
Position
Samsung advocates for practical AI/ML beam management implementations that reuse existing CSI frameworks while introducing minimal spec impact. They push FOR: reusing legacy CSI-ReportConfig structures, supporting both UE-side and NW-side models with clear associated ID mechanisms, and practical performance monitoring with beam accuracy indicators. They push AGAINST: overly complex new signaling frameworks and support maintaining compatibility with existing beam management procedures.
Key proposals
- Proposal 1.1A (Sec 1.1): Updated Option 1 for UE-side model applicability - CSI-ReportConfig provided in Step 3 with associated ID, reported via RRCReconfigurationComplete or UAI, with periodic CSI activated after RRCReconfigurationComplete
- Proposal 2.1-1a (Sec 2.1): Introduce beam accuracy indicator (BAI) for UE-assisted performance monitoring - BAI = Np/N where Np is correct predictions out of N total instances, with options for Top-1/1, 1/Top-K, Top-K/M, and margin-based accuracy
- Proposal 2.2-1b (Sec 2.2): Support dedicated CSI-ReportConfig for monitoring with CSI-ReportConfigID linking to inference configuration, UE measures monitoring resource sets with closest time instance mapping
- Proposal 3.1a (Sec 3.1): For BM-Case 2 reference time, support Option 3 - based on latest transmission occasion of CSI-RS/SSB resource in Set B no later than CSI reference resource
- Proposal 4.1a (Sec 4.1): UE-sided model beam information in inference report is CRI/SSBRI of resource set for Set A, with ranking conveyed by order based on model output probability or predicted RSRPs
- Proposal 5.2 (Sec 5.2): For NW-sided model training data collection, UE reports L1-RSRPs and/or beam-IDs collected by UE as agreed by RAN2
- Proposal 7.1 (Sec 7.1): Study extension of Rel-17 TCI state activation/indication signaling to activate/indicate N joint TCI states for N future time instances in BM-Case2
- Proposal 8.1 (Sec 8.1): Existing CPU mechanism used as starting point for AI/ML-based CSI processing, with FFS on sharing between legacy and AI/ML reporting
- Proposal C (Comments): Send LS to RAN2 on NW-sided model training data collection content - Type 1: all L1-RSRPs, Type 2: Set B L1-RSRPs plus Set A beam info, Type 3: Set B L1-RSRPs plus Set A beam info and RSRPs
- Agreement 118bis (Sec 11.5): RAN1 study three options for UE-side model applicability - Option 1 with CSI-ReportConfig in Step 3, Option 2 with inference parameters, Option 3 with associated ID and UE reporting parameters