R1-2409750
discussion
Specification support for positioning accuracy enhancement
Summary
Tejas Networks Limited submits a contribution discussing AI/ML for positioning accuracy enhancement, focusing on model input definitions, training data collection, and model performance monitoring. The document contains 23 proposals and 15 observations addressing sample-based and path-based measurements, LoS/NLoS indications, and consistency between training and inference phases.
Position
Tejas Networks proposes redefining model outputs for AI/ML positioning by reporting the timing of arrival of the LoS path whether real or virtually inferred, and introducing a new LoS/NLoS indication that characterizes the reported measurement itself rather than the link probability. They propose computing a unique associated ID as a function of TRP-ID, Frequency layer ID, and PRS configurations to ensure consistency between training and inference phases without exposing proprietary spatial filter details. For Model Performance Monitoring, they propose that both metric calculation and decision-making should be centralized at one entity (LMF or UE/TRP) to minimize specification impact, using range error or positioning error as the primary metrics. They propose reusing existing Release-17 frameworks for path-based measurements and existing sampling factors for sample-based measurements to maintain backward compatibility.
Key proposals
- Proposal 1 (Sec Model Input - Sample-based): Supports specific values for the size of the sample window (Nt) based on different deployment scenarios.
- Proposal 2 (Sec Model Input - Sample-based): Proposes that the number of reported samples (Nt') should be a fraction of the window size (Nt), signaled by either the TRP or LMF.
- Proposal 8 (Sec Model Input - Sample-based): Proposes that the start index (n0) in outdoor scenarios should accommodate a wide range due to UEs up to 100 km away.
- Proposal 10 (Sec Model Input - Sample-based): Proposes reusing existing sampling factor (k) values from path-based measurements for sample-based measurements.
- Proposal 11 (Sec Model Input - Path-based): Proposes utilizing the existing Release-17 framework for path-based measurements, potentially increasing the number of reported paths beyond 8.
- Proposal 12 (Sec Model Output): Proposes reporting the timing of arrival of the LoS path, whether real or virtually inferred by the AI-ML model.
- Proposal 13 (Sec Model Output): Proposes a new LoS/NLoS indication that specifies whether the reported measurement itself is LoS/NLoS, rather than the probability of the link being LoS.
- Proposal 14 (Sec Training Data Collection): Proposes a single indicator representing the quality of complete timing information for Case-3b or Case-1.
- Proposal 15 (Sec Training Data Collection): Proposes using both hard (0/1) and soft (0.0-1.0) values for reporting ToA quality and NLoS indication.
- Proposal 17 (Sec Model Inference): Proposes providing specific assistance data to the UE to ensure consistency between training and inference phases.
- Proposal 18 (Sec Model Inference): Proposes computing an associated ID as a function of TRP-ID, Frequency layer ID, PRS configurations, and other parameters to simplify data collection.
- Proposal 21 (Sec Model Performance Monitoring): Proposes considering range error or positioning error as the model performance metric for decision-making.
- Proposal 22 (Sec Model Performance Monitoring): Proposes that both MPM metric calculation and decision-making should occur at a single entity (LMF or UE/TRP) to reduce specification impact.
- Proposal 23 (Sec Model Performance Monitoring): Proposes that AI-ML model performance should be monitored periodically or aperiodically.