R1-2409543
discussion
Discussion on AI/ML for positioning accuracy enhancement
Summary
This document from H3C proposes specific algorithms for eliminating initial phase mismatch in Channel Impulse Response (CIR) measurements used as AI/ML model inputs for NR positioning. It defines reference sample selection criteria (strongest path, first satisfied sample) and reference phase calculation methods (mean, minimum phase), totaling 16 observations and 6 proposals across the document.
Position
H3C supports employing CIR as AI/ML model input for both direct and assistant positioning, arguing it preserves more channel information than PDP or DP. They require the elimination of initial phase mismatch in CIR measurements before AI/ML model input to prevent performance degradation. They prefer the relative phase method over double differential methods to reduce implementation costs and improve algorithm usability. They propose supporting specific reference sample selection criteria, namely 'strongest path' and 'first satisfied sample', and reference phase calculation methods, namely 'mean method' and 'minimum phase method'. They further propose supporting the physical layer process and air interface signaling transmission scheme for LMF-side models to configure and monitor this phase alignment process.
Key proposals
- Proposal 1 (Sec CIR phase mismatch elimination): Defines 'strongest path' as the CIR sample with highest amplitude and 'first satisfied sample' as the first sample exceeding a threshold T (N times noise standard deviation).
- Proposal 2 (Sec CIR phase mismatch elimination): Proposes supporting 'strongest path' and 'first satisfied path' for selecting the 'reference sample' to eliminate initial phase mismatch.
- Proposal 3 (Sec CIR phase mismatch elimination): Defines 'mean method' (averaging phases) and 'minimum phase method' (minimizing sum of imaginary parts after rotation) for calculating 'reference phase'.
- Proposal 4 (Sec CIR phase mismatch elimination): Proposes supporting 'mean method' and 'minimum phase method' for calculating the 'reference phase'.
- Proposal 5 (Sec CIR phase mismatch elimination): Proposes supporting the overall CIR phase mismatch elimination process involving measurement, reference selection, phase calculation, alignment, and monitoring.
- Proposal 6 (Sec Specification impact on phase mismatch elimination): Proposes supporting the physical layer process and air interface signaling transmission scheme for LMF-side models (Case 2b/3b) to handle phase mismatch elimination.
- Observation 1 (Sec CIR as Model Input): Supports CIR measurements as model input for both direct and assistant positioning in AI/ML-based positioning.
- Observation 2 (Sec CIR as Model Input): States that initial phase mismatch of CIR measurements must be eliminated before input to AI/ML models during training and inference.
- Observation 3 (Sec CIR phase mismatch elimination methods): Supports the relative phase method for phase alignment to reduce implementation costs and improve usability compared to double differential methods.
- Observation 4 (Sec CIR phase mismatch elimination methods): Identifies algorithms combining reference sample selection and reference phase calculation as efficient ways to eliminate initial phase mismatch.
- Observation 7 (Sec Conclusions): Reiterates support for CIR measurements as model input for both direct and assistant positioning.
- Observation 8 (Sec Conclusions): Reiterates the necessity of eliminating initial phase mismatch before AI/ML model input.
- Observation 9 (Sec Conclusions): Reiterates support for the relative phase method for phase alignment.
- Observation 16 (Sec Conclusions): Concludes that combining reference sample selection criteria and reference phase calculation methods are efficient and simple ways for eliminating initial phase mismatch.