R1-2410921
discussion
Final summary of specification support for positioning accuracy enhancement
From Ericsson
Ericsson's prior position on
9.1.2
at
RAN1#118bis
· AI-synthesized, paraphrased
verify sources →
Advocates for sample-based measurements over legacy path-based measurements for superior performance and lower complexity, while strongly opposing phase information inclusion in model inputs due to deployment costs and minimal accuracy gains.
Summary
This 3GPP RAN1 document (R1-2410921) from Ericsson presents a final summary of discussions on AI/ML for NR Air Interface positioning accuracy enhancement from RAN1#119 meeting. The document contains over 160 proposals and conclusions across model input/output, training data collection, inference consistency, and model monitoring topics.
Position
Ericsson advocates FOR sample-based measurements as an enhancement to legacy path-based reporting, supporting both alternatives to avoid blocking progress while preferring sample-based for better AI/ML positioning performance. They push AGAINST requiring phase information (CIR) for model input due to implementation complexity and transmitter/receiver circuit variations, and oppose mandatory UTC time reporting. Ericsson supports functionality-level management by LMF while allowing UE-side model monitoring with minimal specification impact.
Key proposals
- Proposal 2.1.1.3-1 (Sec 2.1.1): For sample-based measurement definition, starting time of Nt consecutive samples is determined as: if LMF configures an offset relative to reference time and gNB follows configuration, then starting time = reference time + offset; otherwise starting time = first detected path rounded down with timing granularity T
- Agreement (Sec 2.1.3): For Rel-19 AI/ML based positioning Case 3b, support enhanced sample-based measurement with Nt' values selected from Nt consecutive channel response values, where Nt'≤24, Nt={32,64,128}, timing granularity T=2^k×Tc, and LMF can signal parameters via NRPPa
- Conclusion 2.4.2-1 (Sec 2.4): For Rel-19 AI/ML based positioning for all use cases, there is no consensus in RAN1 to support CIR (Channel Impulse Response) for determining model input
- Conclusion 2.4.2-2 (Sec 2.4): No consensus in RAN1 to support using one phase value (e.g., Rel-18 measurements DL RSCPD, DL RSCP, UL RSCP) for the first path or first sample only for determining model input
- Proposal 3.1.3-1 (Sec 3.1): For AI/ML assisted positioning Case 3a, when LOS/NLOS indicator is reported with legacy timing measurements, the indicator can reuse meaning and format of existing IE 'LoS/NloS Information' in 38.455
- Agreement 3.1.5-1 (Sec 3.1): For measurement report of AI/ML assisted positioning Case 3a, LMF shall be able to distinguish whether timing information is obtained by legacy method or by Rel-19 AI/ML
- Agreement 4.1.4-2 (Sec 4.1): For AI/ML based positioning, when label is UE location and reported, label and quality indicator are provided by reusing existing IE using geographic shapes defined in TS 23.032, with uncertainty and confidence serving as quality indicator
- Agreement 5.1.7-1B (Sec 5.1): For AI/ML based positioning Case 1, all assistance information from legacy UE-based DL-TDOA except info #7 (geographical coordinates of TRPs) can be provided from LMF to UE, with four alternatives for handling info #7 including implicit/explicit provision or associated ID
- Agreement 6.2.4-1 (Sec 6.2): For model performance monitoring of AI/ML positioning Case 1, support Option A where target UE side performs monitoring metric calculation and may signal monitoring outcome to LMF
- Agreement 6.2.5-3 (Sec 6.2): For Case 1 label-based model monitoring, at least Option A-1 (LMF provides ground truth label to UE) and A-2 (LMF provides position calculation assistance data to UE) are supported, with legacy framework and signaling reused