R1-2407797
discussion
Discussion on AI/ML-based positioning enhancement
From ZTE
Summary
ZTE presents a comprehensive technical document on AI/ML-based positioning enhancement for NR air interface with 30 proposals and 8 observations covering model training, inference, monitoring, and data collection aspects for different use cases.
Position
ZTE advocates FOR maximizing reuse of existing 3GPP procedures and specifications rather than defining new enhancements, strongly supports CIR with phase information over PDP for better positioning accuracy despite higher overhead, pushes for sample-based measurements over path-based measurements for unified implementation, and argues AGAINST specifying network-side additional conditions or extra assistance information requirements.
Key proposals
- Proposal 1 (Sec 3.1): From RAN1 perspective, for Case 3b measurements, existing procedures can be reused in terms of SRS configuration for training data collection, monitoring, or inference procedures
- Proposal 2 (Sec 3.2): For AI/ML positioning, different data collection requirements can be configured by the data collection node, with detailed requirements including data size, type, quality, report periodicity
- Proposal 6 (Sec 3.2.3): The association between part A and part B can be determined by the model training entity based on collected measurement information and label information using timestamp and UE ID
- Proposal 12 (Sec 3.3.2): In AI/ML based positioning, support using phase information for determining model input based on CIR performance superiority over PDP
- Proposal 17 (Sec 3.3.4.4): In Rel-19 AI/ML based positioning, regarding time domain channel measurements, support sample-based measurements rather than path-based measurements
- Proposal 18 (Sec 3.4): For AI/ML positioning, additional assistance information of DL PRS configuration is not necessary to be defined for UE side model
- Proposal 19 (Sec 3.5): At least for data collection with LMF-side model, support UE/TRP to report more detailed channel measurements, e.g., CIR
- Proposal 21 (Sec 4.1): For direct AI/ML positioning, no more extra types of model output is required in addition to the output listed in TR 38.843
- Proposal 23 (Sec 4.2): For AI/ML-assisted positioning Case 3a, when LOS/NLOS indicator is reported with legacy timing measurement, reuse existing IE LoS/NLoS Information in 38.455
- Proposal 26 (Sec 5): There is no need to specify any NW-side additional conditions for AI/ML positioning due to sufficient generalization via mixed dataset training
- Proposal 27 (Sec 6.1.1): For LMF-side model, whether assistance information is required for model monitoring depends on whether LMF has prior information on UE/PRU's location
- Proposal 28 (Sec 6.1.2): For model performance monitoring of AI/ML positioning Case 1, support option B where LMF performs monitoring metric calculation rather than UE
- Proposal 29 (Sec 6.2): For AI/ML assisted model monitoring, consider monitoring based on AI/ML inference output (intermediate features) or estimated UE location according to AI/ML inference output