R1-2409845
discussion
Discussion on support for AIML positioning
From InterDigital
Summary
InterDigital's comprehensive technical document presents 38 proposals and 24 observations for AIML positioning in NR, focusing on Case 1 (UE-based positioning with UE-side model), Case 3a (gNB-side model), and Case 3b (LMF-side model). The document prioritizes consistency mechanisms, ground truth quality indicators, and lifecycle management while advocating for path-based measurements over sample-based alternatives.
Position
InterDigital advocates FOR leveraging existing positioning methods (UE-based DL-TDOA) as the foundation for AIML positioning rather than creating new methods, FOR path-based measurements over sample-based alternatives, FOR LMF-controlled ground truth quality indicators, and FOR comprehensive consistency checking between training and inference phases. They push AGAINST introducing new sample-based measurements, AGAINST CIR with all-path phase information, and AGAINST defining new positioning methods when existing ones can be enhanced.
Key proposals
- Proposal 1 (Sec 2.1): Prioritize discussion on determination of consistency between training and inference phase and LCM details for Case 1 and deprioritize discussion related to potentially new measurements
- Proposal 2 (Sec 2.2.1): For Case 1, enhance assistance data used for the existing UE-based DL-TDOA positioning method to support the AIML-based positioning
- Proposal 8 (Sec 2.2.2): Time validity for an AIML model(s) should be specified
- Proposal 12 (Sec 2.2.4.1): In Case 1, the LMF is the only entity that can generate the ground truth label quality indicator associated with location information
- Proposal 14 (Sec 2.2.4.2): For Case 1, support both hard (1 or 0) and soft indicator (0, 0.1, 0.2, …, 1.0) for the ground truth label quality indicator which is generated by the LMF and sent to the training entity
- Proposal 18 (Sec 2.2.5.1): For model performance monitoring of AI/ML positioning Case 1, for model performance monitoring metric calculation in label-based model monitoring, support Option A-1, A-2 and A-3
- Proposal 19 (Sec 2.3.1): For Case 1, enhance the existing UE-based DL-TDOA positioning method to support AIML-based positioning without introducing a new positioning method for AIML-based positioning
- Proposal 24 (Sec 2.4.2.1): For case 3b and case 1, support Alternative (b) (path based measurements) for representation of time domain channel measurements and reuse path-based reporting with the granularity up to k=-6 as specified in Release 18
- Proposal 27 (Sec 2.4.2.1): For direct AI/ML based positioning, for case 1, adopt RSRP, RSRPP and DL-RSTD measurements for the UE to determine input for an AIML model for direct AIML positioning
- Proposal 30 (Sec 2.4.4): Support first-path phase measurement, namely RSCP and RSCPD, for AIML based positioning
- Proposal 32 (Sec 2.5.1): For AIML assisted positioning, support an indication in the measurement report to indicate the reported timing measurement is inferred by an AIML model(s)
- Proposal 36 (Sec 2.6.1): Support multi-RTT positioning method under Case 3b without any specification enhancement
- Proposal 38 (Sec 2.6.2): For Case 3a and Case 3b, use existing UL or DL & UL positioning methods as the starting point for AIML-based positioning method