R1-2500202
discussion
Discussion on AI/ML-based positioning
From CATT
CATT's prior position on
9.1.2
at
RAN1#119
· AI-synthesized, paraphrased
verify sources →
Strongly advocates for sample-based channel measurements over path-based measurements due to superior AI/ML performance and reduced ambiguity. Proposes unified starting time across multiple TRPs using configured or predefined offsets and treats AI/ML positioning as an independent method requiring specialized assistance information. Pushes against separate quality indicators for different measurement components and unnecessary LOS/NLOS reporting when AI/ML derives timing information.
Summary
This document from CATT discusses specification impacts for AI/ML-based positioning in Rel-19, covering data collection, model inference, performance monitoring, and consistency issues across five positioning cases. It contains 37 proposals and 1 observation aimed at defining quality indicators, sample-based measurement reporting, and monitoring mechanisms.
Position
CATT proposes that timing information for a channel measurement be associated with only one quality indicator to reduce overhead, and that LMF provide quality thresholds to filter low-quality training samples. They support sample-based measurements for cases 2b and 3b, specifying candidate values for Nt' (9, 16, 24) and k (0-5), and prefer supporting phase measurement reporting subject to capability. CATT argues against reporting legacy LOS/NLOS indicators when timing is derived from AI/ML, deeming them misleading. For performance monitoring, they support label-based options A-1, A-2, and B-1, and propose that UE report monitoring outcomes including metrics and LCM decisions upon LMF request or condition satisfaction. Finally, they propose using associated IDs to implicitly handle TRP geographical coordinates for consistency and privacy.
Key proposals
- Proposal 1 (Data Collection): Timing information of a channel measurement is only associated with one quality indicator to avoid overhead.
- Proposal 6 (Data Collection): LMF can use a quality indicator condition or criteria to indicate the required quality of collected data for cases 2b and 3b.
- Proposal 8 (Data Collection): The associated measuring time difference between Part A (channel measurement) and Part B (ground truth label) should be restricted to ensure sample validity.
- Proposal 13 (Model Input): Support sample-based measurement for case 2b, composed of Nt' values of estimated channel response, with candidate Nt' values of 9, 16, 24.
- Proposal 15 (Model Input): For case 3b, consider sample-based reporting (using time offset and bitmap) or enhanced path-based reporting for timing information.
- Proposal 19 (Model Input): Phase measurement and reporting is supported for cases 2b and 3b, subject to UE/TRP capability and LMF indication.
- Proposal 22 (Model Output): Do not support reporting legacy LOS/NLOS indicator when timing measurement is obtained by AI/ML, as it may be misleading.
- Proposal 24 (Performance Monitoring): Support Options A-1, A-2, and B-1 for label-based model monitoring metric calculation in cases 1 and 2a.
- Proposal 29 (Performance Monitoring): UE reports monitoring outcome when receiving an LMF request or when pre-configured reporting conditions are satisfied.
- Proposal 30 (Performance Monitoring): Monitoring outcome reports can include monitoring metrics, confidence levels, LCM decisions, requested PRS configuration, and preferred TRP set.
- Proposal 34 (Consistency): Consider two concepts of validity area: assistance data/measurement validity area (AreaID-CellList) and validity area composed of TRPs corresponding to channel measurement.
- Proposal 36 (Consistency): Geographical coordinates of TRPs should be implicitly included in assistance information via an associated ID due to privacy issues.