R1-2410193
discussion
Discussions on AI/ML for beam management
From LG Electronics
Summary
LG Electronics' comprehensive contribution on AI/ML for NR beam management covers data collection, inference, and performance monitoring for both network-sided and UE-sided models, presenting 25 detailed proposals and 4 observations across spatial and temporal beam prediction scenarios.
Position
LG Electronics advocates for practical, overhead-conscious AI/ML beam management solutions, strongly pushing for Alt 4 configuration approach and Set B-only configurations to reduce signaling overhead, while opposing AP CSI-RS support for temporal prediction due to resource concerns. They champion separate AI/ML processing units (APU) distinct from legacy CPU mechanisms and emphasize event-triggered rather than periodic reporting to minimize resource consumption. LG takes a pragmatic stance on beam indication, insisting it should be based on measurable Set B beams rather than potentially unmeasurable Set A beams.
Key proposals
- Proposal 1 (Sec 2.1.1): For NW-sided AI/ML in temporal DL Tx beam prediction, support UE reporting enhancements including past/present best N beam(s) per time stamp and tendency/variance of best N beam(s)
- Proposal 2 (Sec 2.1.2): Support reporting of UE assistance information for determining Set A, e.g., UE to report preferred Set A among candidate beams of Set A
- Proposal 4 (Sec 2.2.1): For more than 4 beam related information in L1 signaling for NW-sided model, support CRI/SSBRI for beam information as legacy with corresponding M L1-RSRP(s)
- Proposal 6 (Sec 2.2.2): For UE-sided model, support Alt 4 for Set A and Set B configuration with one or more separate resource set(s) for Set A configured outside of CSI-ResourceConfig
- Proposal 7 (Sec 2.2.2): Support to configure only resource set for Set B for CSI-ResourceConfig (Alt 1), considering huge resource overhead reduction from not configuring Set A
- Proposal 11 (Sec 2.2.2): Support to report inference results of N(N>=1) future time instance(s) in one report with maximum value of N more than 1
- Proposal 19 (Sec 2.2.2): Support introduction of AI/ML processing units (APU) for CSI reports, occupied for all AI/ML inference operations and separately managed from CPU
- Proposal 20 (Sec 2.3.2): For UE-assisted performance monitoring, support using CSI report configuration for inference with resource set for monitoring same as Set A
- Proposal 21 (Sec 2.3.2): Support event-triggered UE reporting for UE-sided AI/ML performance monitoring via UCI or SR to request changes
- Proposal 22 (Sec 2.4): Confirm working assumption that associated ID can be configured within CSI framework to ensure consistency between training and inference
- Proposal 23 (Sec 2.4): Define similar properties of DL Tx beam set that same downlink spatial domain transmission filters are maintained for each beam in different transmission instances
- Proposal 24 (Sec 2.5): Focus on what information to deliver in each step rather than re-using existing IE for applicable functionality reporting, consider merging Option 1 and Option 2
- Proposal 8 (Sec 2.2.2): Provide assistance information on relation/association between Set A and Set B beams using linear combining coefficients or 2D/3D coordinates while preserving proprietary information
- Proposal 12 (Sec 2.2.2): For BM-Case 2, support differential RSRP report with indicator for time instance containing largest predicted RSRP and differential RSRP values
- Proposal 17 (Sec 2.2.2): For UE-side AI/ML model BM-case 2, do not support AP CSI-RS for Set B configuration due to huge DL resource overhead