R1-2410414
discussion
Discussion on specification support for AI-ML based positioning accuracy enhancement
From Baicells
Summary
Baicells presents their views on AI/ML-based positioning accuracy enhancement for R19, covering model input/output, training data collection, quality indicators, and model monitoring across positioning sub-use cases. The document contains 11 proposals and 9 observations addressing technical aspects from sample-based measurements to model monitoring approaches.
Position
Baicells strongly advocates FOR sample-based measurements over path-based measurements, arguing that sample-based provides superior positioning accuracy (1.06-1.62x better performance) and avoids algorithm inconsistencies between vendors. They push FOR phase information support in Case 3b despite overhead concerns, and advocate FOR minimizing specification impact by reusing existing IEs and procedures wherever possible rather than defining new mechanisms.
Key proposals
- Proposal 1 (Sec 2.1.1): Support sample-based measurements as model input at least for Case 3b. Preclude path-based measurements in R19.
- Proposal 2 (Sec 2.1.2): RAN1 to consider supporting multiple options, such as Option B and Option D for starting time of sample-based measurement.
- Proposal 3 (Sec 2.1.4): Support phase information reporting for Case 3b.
- Proposal 4 (Sec 2.2.1): Reuse existing 'LoS/NLoS information' in TS 38.455 for Case 3a.
- Proposal 5 (Sec 2.4): For Case 1, support the transfer of UE-side model information to LMF via LPP.
- Proposal 6 (Sec 2.5.1): For Case 3a model training, support to explore the feasibility of training models at gNB side as the primary focus of the normative work.
- Proposal 7 (Sec 2.6.1): In NRPPa, the Timing Measurement Quality can be associated with the timing information of Part A for each TRP, which can later be transferred to LMF.
- Proposal 8 (Sec 2.6.2): For Case 3a and 3b inference, support reusing existing IE 'Timing Measurement Quality' to align the measurement for both training and inference with traditional positioning methods.
- Proposal 9 (Sec 2.6.2): There is no strong need for power quality indicator considering it's only for PDP and has not been well evaluated in the past.
- Proposal 10 (Sec 2.6.3): For quality indicator of Case 1 label, support Option 1, reusing an existing IE.
- Proposal 11 (Sec 2.7.1): Support Option B-1 for Case 1, which can be achieved with minimum specification effort.