R1-2500618
LS out
[Draft] Reply LS on LMF-based AI/ML Positioning for Case 2b
From ZTE
Summary
This document is a Liaison Statement reply from RAN1 to SA2 regarding LMF-based AI/ML Positioning for Case 2b in Release 19. It outlines current agreements on data types (timing, power, phase) and measurement alternatives (sample-based vs. path-based), as well as procedures for model training data collection and performance monitoring. The document contains no new proposals but summarizes existing RAN1 agreements and future work plans across data types, training procedures, and monitoring.
Position
ZTE, representing RAN1, confirms that for AI/ML positioning Case 2b, the supported data types include timing information and paired timing/power information, with phase information still under investigation. They propose studying two alternatives for time domain channel measurements: sample-based measurements (defined by Nt' samples with granularity T=2kxTc) and path-based measurements (based on existing Rel-18 reporting). For model training, they define data samples comprising Part A (channel measurements with NR-TimingQuality indicators) and Part B (ground truth labels), assuming these parts correspond to the same UE and location. They require that power measurements use DL PRS-RSRPP and UL SRS-RSRPP as starting points. Regarding model monitoring, they propose studying the necessity of assistance information from UE/PRU/gNB to LMF, while noting that monitoring metric calculation is out of RAN1 scope.
Key proposals
- Agreement RAN1#116 (Sec 1): Supports timing information and paired timing/power information as minimum data types for UE reporting to LMF in Case 2b.
- Agreement RAN1#116 (Sec 1): Investigates sample-based measurements (integer multiple of sampling periods) and path-based measurements (detected path timing) for time domain channel measurements.
- Agreement RAN1#116 (Sec 1): Investigates the necessity and feasibility of using phase information (e.g., for first path or CIR triplet) in addition to timing and power.
- Agreement RAN1#117 (Sec 1): Defines sample-based measurement as Nt' samples of estimated channel response with timing granularity T=2kxTc, where Nt' and k can be signaled.
- Agreement RAN1#117 (Sec 1): Defines path-based measurement as existing Rel-18 measurement reporting with potential enhancements on the number of reported paths.
- Agreement RAN1#118 (Sec 1): For sample-based measurement, selects Nt' samples out of Nt consecutive samples, expecting TRP/gNB to select those with highest power.
- Agreement RAN1#118 (Sec 1): Uses DL PRS-RSRPP and UL SRS-RSRPP defined in TS 38.215 as starting points for downlink and uplink power measurements, respectively.
- Agreement RAN1#116b (Sec 1): Defines training data samples as including Part A (channel measurement, quality indicator, timestamp) and Part B (ground truth label, quality indicator, timestamp).
- Working Assumption RAN1#116b (Sec 1): Assumes channel measurement data is generated by PRU and/or non-PRU UE, while labels can be generated by PRU, non-PRU UE with estimated location, or LMF.
- Agreement RAN1#117 (Sec 1): Assumes that if a training sample contains both Part A and Part B, they correspond to the same UE and the same location associated with Part B.
- Agreement RAN1#118b (Sec 1): Reuses existing NR-TimingQuality IE from TS 37.355 and TS 38.455 for the quality indicator of timing information in Part A.
- Agreement RAN1#116 (Sec 1): Studies whether/what assistance information or measurement reports from UE/PRU/gNB to LMF are needed for performance monitoring of LMF-side models.
- Future Work Plan (Sec 1): Plans to discuss detailed signaling of timestamps and quality of channel measurements for Case 2b.
- Future Work Plan (Sec 1): Plans to discuss details of measurements reported by UE/PRU to LMF for model inference.