R1-2410101
discussion
On specification for AI/ML-based beam management
From OPPO
Summary
OPPO's technical document on AI/ML-based beam management for 3GPP RAN1 contains comprehensive responses to RAN2 liaison questions and presents 50+ technical proposals covering NW-side and UE-side model inference, training, monitoring, and consistency mechanisms. The document addresses both BM-Case1 (spatial-domain) and BM-Case2 (temporal-domain) beam prediction scenarios.
Position
OPPO strongly advocates for leveraging existing CSI framework for AI/ML beam management configuration and supports UE-side model autonomy through Type 2 performance monitoring where UE makes LCM decisions. They push against requiring UE-side additional conditions for NW-side models and favor probability-based performance metrics over L1-RSRP differences. OPPO promotes separate CPU counting for AI/ML vs legacy CSI reporting and emphasizes overhead reduction through implicit reporting mechanisms.
Key proposals
- Proposal 1 (Sec 3.1.2): For BM-Case2 with NW-side model, support that UE can be indicated with multiple TCI states for multiple future instances
- Proposal 2 (Sec 3.1.3): For NW-side model training, UE reports L1-RSRPs measurements of fixed Set B as model inputs and Top-K L1-RSRP(s) and Top-1 Tx beam index as labels
- Proposal 3 (Sec 3.1.4): For BM-Case1 and BM-Case2 with NW-side model, it is NOT necessary to specify UE-side additional condition on UE Rx beamforming
- Proposal 4 (Sec 3.2.1.1): For inference with UE-side model, support UE to report beam information on predicted Top K beam(s) and probability information of predicted Top K beam(s)
- Proposal 5 (Sec 3.2.1.2): For inference of BM-Case2 with UE-side model, support to enhance the differential RSRP reporting with UE packing predicted beams and RSRPs from 1st to N-th time instances
- Proposal 6 (Sec 3.2.1.3): For BM-Case2 with UE-side model, support to configure two resource sets for Set A and Set B separately in a CSI-ReportConfig
- Proposal 7 (Sec 3.2.2.1): For performance monitoring with UE-side model, support Type 2 (indication/request/report from UE to gNB) performance monitoring
- Proposal 8 (Sec 3.2.2.2): For Type 1 Option 2 performance monitoring, support the probability of model output (Alt.4) as one performance metric
- Proposal 9 (Sec 3.2.2.4): For monitoring Type 1 Option 2 with UE-side model, support separate resource set configuration in a dedicated CSI report configuration
- Proposal 10 (Sec 3.2.3): For inference with UE-side model, support that each associated ID corresponds to a Set B and Set A as a triple of {associated ID, Set B, Set A} configured within CSI framework
- Proposal 11 (Sec 3.2.4): For data collection of UE-side model training, support to configure Set B (as model inputs) and Set A (as model labels) using existing CSI framework
- Proposal 12 (Sec 3.3): For BM-Case1 and BM-Case2 with UE-side model, the overall CPU should be separately counted between legacy CSI reporting and AI/ML-based CSI reporting
- Proposed Answer 4-1 (Sec 2): The associated ID(s) can be optionally configured by NW in Step 3 to represent NW-side additional conditions
- Proposed Answer 6 (Sec 2): The inference resource sets for Set B and Set A are configured within CSI framework, i.e. CSI-ReportConfig, in Step 5
- Proposal 13 (Sec 3.3): For BM-Case2, suggest studying and evaluating the beam dwelling time prediction