R1-2410717
discussion
Summary #4 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 (Tdoc R1-2410717) from Ericsson summarizes discussions on AI/ML-based positioning accuracy enhancements from RAN1#119, containing over 80 proposals across model input definitions, output specifications, training data collection, and model inference procedures. The document focuses on technical specifications for sample-based vs path-based measurements, LOS/NLOS indicators, and lifecycle management for AI/ML positioning cases.
Position
Ericsson advocates FOR sample-based measurements over path-based measurements, supporting flexible parameter configurations (Nt, Nt', k) while maintaining backward compatibility with legacy positioning methods. They push FOR reusing existing IEs and frameworks to minimize specification impact, and advocate AGAINST supporting CIR measurements due to complexity concerns. Ericsson supports UE-side monitoring (Option A) over LMF-side monitoring for Case 1, emphasizing implementation flexibility while ensuring consistency between training and inference phases.
Key proposals
- Proposal 2.1.1.2-1 (Sec 2.1.1): For sample-based measurement, starting time of Nt consecutive samples is the timing of first detected sample within a search window, with starting point determined by configured offset relative to reference time
- Proposal 2.1.3.4-1A (Sec 2.1.3): Support sample-based measurements with Nt' values selected from Nt consecutive channel response values, where Nt'={9,16,[24],[32]}, Nt={32,64,128}, k={3,4,5}
- Conclusion 2.4.2-1 (Sec 2.4): No consensus in RAN1 to support CIR (Channel Impulse Response) for determining model input in Rel-19 AI/ML based positioning
- Proposal 3.1.2-1A (Sec 3.1): LOS/NLOS indicator for AI/ML assisted positioning reuses meaning from TS 38.455 and can reuse existing IE 'LoS/NloS Information' format
- Proposal 4.1.4-2 (Sec 4.1): For training data collection, when label is UE location, reuse existing IE using geographic shapes defined in TS 23.032 with uncertainty and confidence as quality indicators
- Proposal 4.4.2-1 (Sec 4.4): For training data collection, if Part A and Part B generated by same entity, provide paired data in one report sharing same time stamp
- Proposal 5.1.6-2 (Sec 5.1): For AI/ML positioning Case 1, all assistance information from legacy UE-based DL-TDOA may be provided explicitly, with geographical coordinates provided either implicitly or explicitly via associated ID
- Agreement (Sec 6.2): For model performance monitoring Case 1, support Option A where target UE performs monitoring metric calculation and may signal monitoring outcome to LMF
- Proposal 6.2.5-3 (Sec 6.2): Support at least Option A-1 and A-2 for label-based monitoring, reusing legacy framework, measurements and signaling
- Agreement (Sample-based): Starting time of Nt consecutive samples determined as first detected path rounded down with timing granularity T for gNB/TRP measurements