R1-2409478
discussion
Discussion and reply LS on applicable functionality reporting for beam management UE-sided model
From ZTE
ZTE's prior position on
5
at
RAN1#118bis
· AI-synthesized, paraphrased
verify sources →
Advocates for flexible and optional network-side signaling to minimize overhead while supporting sub-use-case level granularity and reusing existing CSI framework mechanisms, opposing mandatory associated ID signaling.
Summary
ZTE analyzes the applicable functionality reporting procedures for UE-sided AI/ML beam management models, specifically addressing questions from a RAN2 Letter of Agreement. The document presents 15 distinct proposals across four sections, arguing against Option 1 for inference applicability due to resource waste and scheduling complexity, while strongly supporting Option 3 for its compatibility with legacy CSI report designs.
Position
ZTE opposes Option 1 for UE-side model inference applicability, arguing it causes substantial waste of air interface resources and complicates NW scheduling. They support Option 3, emphasizing its compatibility with legacy CSI-ReportConfig design. ZTE proposes that AI/ML beam management feature groups be defined at least per sub-use case level (BM-Case1, BM-Case2), with condition combinations aligned between NW and UE outside UE capability indication to support future extensions. They argue that the provision of associated ID in Step 3 should be optional and that inference configuration in Step 3 is unnecessary for determining applicability. ZTE requires that Step 4 reporting include condition combinations and model availability, while Step 5 inference configuration comprises associated CSI report configurations. Finally, they propose that functionality activation depends on CSI report temporal characteristics and that L1/L2 signaling for activation/deactivation should reuse existing CSI report mechanisms.
Key proposals
- Proposal 1 (Discussion about three options): Opposes Option 1 for inference applicability, citing substantial waste of air interface resources, unclear timing for CSI report activation, and complicated NW scheduling procedures.
- Proposal 2 (Discussion about three options): Supports Option 3 for inference applicability due to its compatibility with legacy CSI report design and facilitation of easier NW scheduling.
- Proposal 3 (On general part): Proposes that for AI/ML beam management, the feature group is defined at least per sub-use case level (e.g., BM-Case1, BM-Case2), with component values signaled via UE capability indication in Step 2.
- Proposal 4 (On general part): Proposes that condition combinations supported by the UE are aligned between NW and UE outside the UE capability indication to ensure compatibility with future AI/ML technology extensions.
- Proposal 5 (On general part): Clarifies that the granularity of functionality, expected to be at least at the sub-use case level or finer, is not solely determined by the UE capability indication in Step 2.
- Proposal 6 (On NW-side additional condition and configuration part): Proposes that a single functionality might be associated with one or multiple associated IDs, and conversely, different functionalities might share the same associated ID.
- Proposal 7 (On NW-side additional condition and configuration part): Proposes that the provision of associated ID in Step 3 is optional, representing an implementation behavior of the NW side.
- Proposal 8 (On NW-side additional condition and configuration part): Proposes that the UE can train and recommend applicable functionalities without the signaling of associated ID, while the final decision on functionality applicability remains with the network.
- Proposal 9 (On NW-side additional condition and configuration part): Proposes that providing configuration (e.g., inference configuration) other than NW-side additional condition in Step 3 is not necessary to avoid unnecessary signaling consumption.
- Proposal 10 (On NW-side additional condition and configuration part): Proposes that the content of applicable functionality reporting in Step 4 includes condition combinations (sets of inference-related parameters) with model availability and applicable associated ID.
- Proposal 11 (On NW-side additional condition and configuration part): Proposes that the content of inference configuration in Step 5 comprises configurations of the associated CSI report.
- Proposal 12 (On Functionality activation part): Proposes that if inference configuration is provided in Step 3, it does not activate the functionality immediately upon receiving Step 3.
- Proposal 13 (On Functionality activation part): Proposes that if inference configuration is not provided in Step 3, configuration in Step 5 does not necessarily activate the functionality immediately, depending on the temporal characteristics of the associated CSI report.
- Proposal 14 (On Functionality activation part): Proposes that if more than one functionality is configured in Step 5, multiple/all applicable functionalities can be activated if simultaneous processing does not exceed UE capabilities.
- Proposal 15 (On Functionality activation part): Proposes that L1/L2 signaling for functionality activation/deactivation is needed, which can reuse the activation/deactivation signaling for CSI report.