R1-2409751
discussion
Discussion on study for AI/ML CSI prediction
Summary
Tejas Networks discusses AI/ML for CSI prediction in Rel-19, focusing on ensuring consistency between training and inference, Life Cycle Management (LCM) modes, data collection mechanisms, and performance monitoring strategies. The document presents 14 proposals and 3 observations addressing issues such as interference variations, model identification, CSI-RS configuration, and threshold-based reporting.
Position
Tejas Networks proposes using an associated ID to align training and inference conditions, mitigating performance degradation from interference variations and TxRU mapping differences. They support AI/ML model identification in LCM mode, where the Network identifies models by Model ID and the UE indicates support. For data collection, they propose reusing Rel-18 Type II Doppler codebook CSI-RS configurations and allowing the UE to report the minimum required CSI-RS instances. Regarding performance monitoring, they prioritize Type 1 (Network-configured) monitoring, proposing that the Network assigns threshold criteria and the UE reports monitoring output only when it falls below these thresholds to reduce signaling overhead. They define specific specification impacts for monitoring, including model switching, parameter updates, and fallback operations based on metrics like SGCS and NMSE.
Key proposals
- Proposal 1 (Consistency): Use an associated ID-based solution to ensure consistency between training and inference phases, extending methods from AI-based beam management.
- Proposal 2 (Consistency): Consider three types of performance monitoring-based solutions: Type 1 (Network-configured), Type 2 (UE-based), and Type 3 (Joint network and UE-based).
- Proposal 3 (LCM mode): Support AI/ML model identification where models are identified by Model ID by the Network, and the UE indicates supported AI/ML models.
- Proposal 4 (Data collection): Consider options for initiating CSI-RS transmission for prediction, including NW-triggered or UE-initiated transmission.
- Proposal 5 (Data collection): Consider that CSI-RS transmission for CSI prediction can be periodic or aperiodic.
- Proposal 6 (Data collection): Consider that the UE should report the minimum number of CSI-RS instances required for CSI prediction to the Network.
- Proposal 7 (Data collection): Reuse the CSI-RS configuration from Rel-18 Type II Doppler codebook for model inference.
- Proposal 8 (Data collection): Consider defining the maximum time gap allowed between the prediction and observation window to improve prediction accuracy.
- Proposal 9 (Performance monitoring): Define specification impacts for performance monitoring, including CSI-RS configuration, threshold criteria, measurement, reporting, model switching, and fallback operations.
- Proposal 10 (Performance monitoring): Consider that the Network assigns the threshold criterion for the UE.
- Proposal 11 (Performance monitoring): Consider that the UE only reports performance monitoring output when the monitoring output is below the threshold.
- Proposal 12 (Performance monitoring): Define specification impacts for Type 3 monitoring, including metric measurement, reporting to UE, model switching, and fallback.
- Proposal 13 (Performance monitoring): For Type 1 and 3 monitoring, use intermediate KPIs (e.g., SGCS, NMSE) or statistical values (e.g., mean SGCS, NMMSE) as performance metrics.
- Proposal 14 (Performance monitoring): Prioritize Type 1 performance monitoring.