R1-2409926
discussion
Specification support for AI/ML-based positioning
From CATT
Summary
CATT's comprehensive technical document presents 35 proposals and 7 observations for AI/ML-based positioning across the NR air interface, covering data collection, model inference, performance monitoring, and consistency issues for all positioning cases (1, 2a, 2b, 3a, 3b).
Position
CATT strongly advocates FOR sample-based channel measurements over path-based measurements due to superior AI/ML performance and reduced ambiguity, unified starting time across multiple TRPs using configured/predefined offsets rather than implementation-dependent methods, and treating AI/ML positioning as an independent method requiring specialized assistance information. They push AGAINST separate quality indicators for different measurement components and unnecessary LOS/NLOS reporting when AI/ML derives timing information.
Key proposals
- Proposal 1 (Sec 2.1.1.1): For training data collection of AI/ML based positioning, timing information of a channel measurement is only associated with one quality indicator
- Proposal 6 (Sec 2.2.1): A quality threshold can be provided from LMF to channel measurement/label generation entity to control data quality and reduce overhead
- Proposal 12 (Sec 3.2): At least for case 3b and 2b, support sample-based channel measurements as the AI/ML model input
- Proposal 13 (Sec 3.3.1): The starting point of the Nt consecutive samples is determined by a configured or predefined offset
- Proposal 18 (Sec 3.3.6): Phase measurement and reporting is supported for case 2b and case 3b, up to UE/TRP capability
- Proposal 21 (Sec 3.4): For AI/ML-assisted positioning, no need to report LOS/NLOS indicator when timing information is reported based on the output of AI/ML model
- Proposal 23 (Sec 4): For label-free monitoring, for all cases, the performance metric calculation is up to implementation
- Proposal 29 (Sec 5.1.1): For AI/ML-based positioning, both concepts of validity area for training data collection can be considered: assistance data validity area and TRP-based validity area
- Proposal 31 (Sec 5.3): AI/ML based positioning should be considered as an independent positioning method, and the assistance information needs to be re-considered
- Proposal 14 (Sec 3.3.2): Candidate values for Nt include at least 16, 32, 64, 128; Nt' include at least 9, 16, 25, 32; k include at least 0…5
- Proposal 3 (Sec 2.1.2): For case 3a, ground truth label generated by LMF considers UE/PRU location information or multiple gNB/TRP SRS-pos measurements
- Proposal 8 (Sec 2.2.2): Time stamp definitions for Part A (channel measurement time) and Part B (label validity time, SRS-pos send time, or PRS measurement time)
- Proposal 24 (Sec 4): For case 1 and 2a model monitoring, support Options A-1, A-2 and B-1 for ground truth label generation
- Proposal 33 (Sec 5.3): At least PRS configuration, validity area, beam information, and synchronize errors should be explicitly included in the assistance information
- Proposal 35 (Sec 6): On AI/ML functionality management for Case 1 and Case 2a, wait for more progress of applicable functionality reporting in AI/ML-based beam management