R1-2410571
discussion
Discussion on specification support for AI/ML Positioning Accuracy enhancement
From CEWiT
Summary
CEWiT presents 18 proposals for AI/ML-based positioning accuracy enhancement, covering sample-based measurement reporting, training data collection procedures, and model monitoring frameworks across different positioning use cases. The document addresses time domain channel measurements, quality indicators, timing information handling, and model lifecycle management for UE-sided and network-sided AI/ML positioning models.
Position
CEWiT advocates FOR sample-based over path-based channel measurement reporting due to better positioning accuracy despite higher overhead, supports distributed model monitoring responsibilities (gNB for Case 3a, LMF for Cases 2b/3b), and pushes for semi-supervised learning to leverage both PRU and non-PRU UEs for training data collection. They are implicitly arguing AGAINST centralized model management approaches and promoting UE autonomy in parameter selection while maintaining network assistance.
Key proposals
- Proposal 1 (Sec 2.1): In Rel-19 AI/ML based positioning, regarding the time domain channel measurements, time domain sample-based measurement reporting is preferred over path-based reporting
- Proposal 2 (Sec 2.1): Based on the UE capability report and scenario related information, in Case 2b, the LMF can recommend sample selection criteria, optimal ranges for Nt' and k to UE
- Proposal 4 (Sec 2.1): Support considering time at which the first path is detected with respect to a reference time as the starting of the sampling window for model input data
- Proposal 6 (Sec 2.2.1): Channel Measurements can be performed by both PRU and non PRU UE
- Proposal 7 (Sec 2.2.1): Semi supervised learning can be used for positioning Non PRU UE in scenarios where lesser PRUs are deployed in an area
- Proposal 8 (Sec 2.2.1): To associate the measurement and its related data to collected labels, the data generating entity can record its ID (UE ID/TRP ID/LMF ID), RS configurations and corresponding IDs, Area ID along with collected measurement, labels and corresponding time stamps
- Proposal 9 (Sec 2.2.2): Support reporting UTC time to LMF everytime SFN completes its cycle
- Proposal 10 (Sec 2.2.3): For all use cases (Case 1, Case2a, Case 2b, Case 3a, Case 3b), RS configuration, related UE and gNB associations and Area validity from which training data is collected should be provided as assistance information to the data collecting entity
- Proposal 12 (Sec 2.2.4): For reporting measurement power quality, SNR relative to a predefined threshold to be specified
- Proposal 14 (Sec 2.3.1): For label free monitoring for Case 3a, gNB/TRP is preferred to perform the model monitoring metric calculation
- Proposal 15 (Sec 2.3.1): For label-based monitoring for Case 3a, LMF is preferred to perform the model monitoring metric calculation
- Proposal 16 (Sec 2.3.1): In case 3a, gNB is preferred to make model management decisions for its own model and notifies its decision to LMF
- Proposal 18 (Sec 2.3.2): LMF is preferred to take model management decisions for Case 2b and Case 3b