R1-2410360
discussion
Discussion on specification support for AI/ML based positioning accuracy enhancements
From Sharp
Summary
Sharp's technical document discusses specification support for AI/ML based positioning accuracy enhancements, covering sample-based measurements, model input/output, consistency between training and inference, and performance monitoring. The document contains 9 proposals and 10 observations addressing various aspects of AI/ML positioning implementation in NR.
Position
Sharp advocates for flexible, implementation-friendly approaches to AI/ML positioning. They push FOR supporting multiple options simultaneously rather than down-selecting (Options A, C, D can co-exist), continuous phase information over single-phase, and Option 1-1 interpretation for LOS/NLOS indicators. They are AGAINST restricting implementations to single approaches and prefer solutions that reduce signaling overhead while maintaining consistency between training and inference phases.
Key proposals
- Proposal 1 (Sec 1.1.1.1): For sample-based measurement, study Options A, C, and D for determining starting time of Nt consecutive samples - Option A uses first detected path timing, Option C uses fixed offset, Option D uses configured offset
- Proposal 2 (Sec 1.1.2): For AI/ML based positioning, both Legacy-like format and Bitmap format should be supported for channel measurement in sample-based measurement
- Proposal 3 (Sec 1.1.3): For AI/ML based positioning, continuous phase information is supported for model input rather than single-phase information
- Proposal 4 (Sec 1.2.1): For AI/ML assisted positioning case 3a, study options for LOS/NLOS indicator - Option 1 provides indicator with spec change, Option 2 does not provide indicator when timing is AI/ML predicted
- Proposal 5 (Sec 1.2.1): For AI/ML predicted timing information, the LOS/NLOS indicator provides confidence of virtual LOS propagation path of the AI/ML model output
- Proposal 6 (Sec 1.2.2): For AI/ML assisted positioning Case 3a, report an indicator that provides whether the timing information is made by AI/ML
- Proposal 7 (Sec 2.1): For AI/ML positioning Case 1, explicitly signal assistance data including PCIs, GCIs, ARFCN, PRS IDs, DL-PRS configuration, spatial direction info, and TRP geographical coordinates to ensure consistency between training and inference
- Proposal 8 (Sec 3.1): For performance monitoring in AI/ML positioning Case 1, support Option A-3 when PRU is involved in performance monitoring
- Proposal 9 (Sec 3.1): For performance monitoring in AI/ML positioning Case 1, support Option A-2 when PRU is not involved in performance monitoring