R1-2500614
LS out
Draft LS reply on LMF-based positioning for case 3b
From NEC
Summary
This document serves as a Liaison Statement reply from RAN1 to SA2 regarding the progress of AI/ML-based positioning for Case 3b in Release 19. It consolidates existing RAN1 agreements and working assumptions concerning data collection, specifically defining the structure of training data into Part A (channel measurements) and Part B (ground truth labels), and identifies the entities responsible for generating these components.
Position
NEC presents the consolidated RAN1 agreements on AI/ML positioning data collection for Case 3b, proposing that training data be structured into Part A (channel measurements) and Part B (ground truth labels). They require the reuse of existing IEs, specifically NR-TimingQuality in TS 37.355 and Timing Measurement Quality in TS 38.455, for reporting timing quality indicators. NEC supports the assumption that Part A is generated by TRP/gNB, while Part B can be generated by PRU, Non-PRU UE, or LMF. They maintain the working assumption that Part A and Part B within a single training sample correspond to the same UE and location. NEC notes that the detailed definition of channel measurement and the specific association of quality indicators to timing information remain FFS items.
Key proposals
- Proposal 1 (Sec 1): RAN1 reports that for Case 3b, at least timing information or paired timing and power information can be reported as input data.
- Proposal 2 (Sec 1): Training data is structured into Part A (channel measurement, quality indicator, timestamp) and Part B (ground truth label, quality indicator, timestamp).
- Proposal 3 (Sec 1): Existing IEs such as NR-TimingQuality in TS 37.355 and Timing Measurement Quality in TS 38.455 should be reused for timing quality indicators.
- Proposal 4 (Sec 1): The existing IE 'Time Stamp' in TS 38.455 can be reused for channel measurement timestamps.
- Proposal 5 (Sec 1): Part A and Part B in a single training data sample are assumed to correspond to the same UE and the same location associated with Part B.
- Proposal 6 (Sec 1): Part A (measurements) is agreed to be generated by TRP/gNB.
- Proposal 7 (Sec 1): Part B (labels) is assumed to be generated by PRU, Non-PRU UE with estimated location, or LMF.
- Proposal 8 (Agreement @RAN1#116): Supports reporting of time domain channel measurements including timing information and paired timing/power information.
- Proposal 9 (Agreement @RAN1#116b): Defines the components of Part A and Part B for training data collection, noting that detailed definition of channel measurement is FFS.
- Proposal 10 (Agreement @RAN1#116b): Confirms that measurement and related data (e.g., timestamp) for Case 3a and 3b are generated by TRP/gNB.
- Proposal 11 (Working Assumption @RAN1#117): Assumes label generation entities for Case 3b include PRU, Non-PRU UE, or LMF, with transfer out of RAN1 scope.
- Proposal 12 (Agreement @RAN1#117): Assumes Part A and Part B in one sample are for the same UE and location, with association details to be discussed.
- Proposal 13 (Agreement @RAN1#118/118b): Specifies using existing NR-TimingQuality IEs for quality indicators, with association details marked as FFS.