R1-2410215
discussion
Discussion on specification support for positioning accuracy enhancement
From TCL
Summary
TCL presents their position on AI/ML based positioning for NR air interface, covering performance monitoring, training data collection, and consistency between training and inference. The document contains 11 proposals and 2 observations addressing various aspects of AI/ML positioning implementation.
Position
TCL advocates FOR Option A-2 for label-based model monitoring to reduce data transfer overhead, supports AI-specific reference signal configurations with hierarchical resource type structures, and promotes explicit indication of critical assistance data IEs for consistency. They push FOR UE-side validation in training data association and configurable offset approaches for sample timing determination, while arguing AGAINST high-overhead options that require multiple measurement data transfers.
Key proposals
- Proposal 1 (Performance monitoring): For model performance monitoring metric calculation in label-based model monitoring for AI/ML positioning Case 1, Option A-2 is preferred
- Proposal 2 (Performance monitoring): It is recommended to perform the monitoring metric calculation at the model inference entity for label-free model monitoring
- Proposal 3 (Performance monitoring): LMF is responsible for functionality management decision making
- Proposal 5 (Sample-based measurement): Down-select options for determining starting time of Nt consecutive samples using fixed offset (Option C) or configured offset (Option D) relative to reference time
- Proposal 6 (Training data collection): For Case 3b, assistance information for training data association should be discussed when models are trained at LMF and Part B is generated by UE
- Proposal 7 (Training data collection): Down-select methods for training data association - UE validation with availability indicator or LMF validation with immobility duration
- Proposal 8 (AI-enabled reference signals): AI-specific or AI model-specific reference signal configurations, including PRS and SRS for positioning, could be introduced and indicated to UE
- Proposal 9 (Additional conditions): Assistance data IEs used for channel estimation and those not impacting consistency should be explicitly indicated, others depend on impact sensitivity
- Proposal 11 (Additional conditions): Associated ID can be adopted for consistency between training and inference for Case 1, with FFS on format of associated IDs
- Observation 1 (AI-enabled reference signals): Conventional reference signal configurations could be enhanced to accommodate AI-specific requirements
- Observation 2 (Additional conditions): There should be no specification impact regarding additional conditions for Case 3a