R1-2500338
discussion
Specification support for positioning accuracy enhancement
From vivo
Summary
vivo presents a comprehensive contribution for AI/ML-based positioning in NR, focusing on data collection, model inference, and monitoring for Cases 1, 2a, 2b, 3a, and 3b. The document contains 36 proposals and 9 observations, advocating for the extension of agreements from 1st priority cases to 2nd priority cases, the reuse of legacy IEs for quality indicators, and specific parameter sets for sample-based channel measurements.
Position
vivo proposes extending agreements from 1st priority cases to 2nd priority cases, specifically supporting sample-based channel measurements for Case 2b. They require reusing existing IEs for quality indicators, opposing the definition of new abstract IEs for label or channel measurement quality. vivo argues against mandating UTC time alongside SFN for Case 1, stating LMF implementation can resolve SFN wrapping ambiguity. They propose specific candidate sets for Nt' ({8, 16, 24}) and k ({0, 1, 2, 3, 4, 5}) for sample-based reporting, and oppose transmitting offset from gNB to LMF when parameters differ from recommendations. vivo supports using phase information (RSCP/RSCPD) to enhance positioning accuracy and model monitoring, proposing methods like PRU-assisted calibration or relative phase referencing to mitigate transceiver initial phase impacts. They require supporting distance ranging for Cases 2a/3a and prioritizing functionality terminology discussions to align with RAN2.
Key proposals
- Proposal 1 (Sec 2.1): Resume the discussion of the 2nd priority cases, noting that agreements from 1st priority cases can be extended with minor changes.
- Proposal 2 (Sec 3.1): Associate Part A and Part B for training data samples using timestamps at model training entities or by reporting paired data from the same generation entity.
- Proposal 4 (Sec 3.2): Do not support always mandatorily reporting UTC time together with SFN for Case 1 data collection, as ambiguity can be avoided by LMF implementation.
- Proposal 8 (Sec 4.1.1): Define candidate Nt' values as {8, 16, 24} for sample-based channel measurements, where UE/gNB reports at least Nt' samples within Nt consecutive samples.
- Proposal 12 (Sec 4.1.2): Do not support transmitting offset from gNB to LMF when used parameters differ from LMF recommendations, arguing it is unnecessary.
- Proposal 14 (Sec 4.1.3): For Case 3b, LMF should indicate whether sample-based channel measurement is requested through a new indicator.
- Proposal 21 (Sec 4.1.4): For Case 3b, the reported channel sample list should comprise Nt' samples including timing, power, and optionally phase information.
- Proposal 26 (Sec 4.1.5): Reuse legacy reference time for UE side channel measurements, specifically TSubframeRxi or TUE-TX.
- Proposal 29 (Sec 4.1.6): Support raw CIR reporting, allowing LMF to eliminate the impact of random initial phases of transceivers by implementation.
- Proposal 30 (Sec 4.2): Support distance ranging between UE and TRP for reporting in AI/ML assisted positioning Cases 2a and 3a.
- Proposal 32 (Sec 5.1): Support Options A-1, A-2, and A-3 for model performance monitoring of AI/ML based positioning Case 1.
- Proposal 34 (Sec 5.2): Define monitoring outcome content to include an indicator of whether the current AI/ML functionality is valid or deactivated.
- Proposal 35 (Sec 6): Prioritize discussion of functionality concepts/terminology for Case 1 to align with RAN2 procedures.
- Proposal 36 (Sec 7): Prefer alternatives other than Alternative 3 for providing info #7 from LMF to UE for Case 1.