R1-2500547
discussion
AI/ML based CSI Prediction
From Google
Google's prior position on
9.1.3
at
RAN1#119
· AI-synthesized, paraphrased
verify sources →
Advocates for UE-side only CSI prediction models with UE-reported preferred CSI-RS configurations to avoid configuration mismatches. Proposes associated ID management per CSI report configuration to maintain training/inference consistency. Pushes for dynamic activation/deactivation capabilities to enable network energy saving while maintaining model performance.
Summary
Google proposes reusing the inference configuration and performance monitoring frameworks established for AI/ML-based Beam Management (BM) to ensure consistency for AI/ML-based CSI prediction. The document outlines six specific proposals covering configuration reuse, CQI offset metrics, separate CSI-RS resource sets, and NW-triggered reporting via UCI.
Position
Google supports reusing the inference configuration procedure for AI/ML based BM for AI/ML based CSI prediction to maintain consistency of training and inference. They support reusing the framework for performance monitoring configuration, specifically requiring the NW to configure one CSI report configuration for inference and another linked configuration for monitoring. Google supports using the CQI offset between predicted and ground-truth CSI as the performance monitoring metric. They support the NW configuring separate CSI-RS resource sets for inference and monitoring, applicable to periodic, semi-persistent, or aperiodic CSI-RS. Google supports NW-triggered performance monitoring results reports and deprioritizes event-triggered reports. Finally, they support reporting performance monitoring results via UCI to manage latency.
Key proposals
- Proposal 1 (Consistency of training and inference): Support reusing the inference configuration procedure for AI/ML based BM for AI/ML based CSI prediction to maintain consistency of training and inference regarding supported/applicable functionalities.
- Proposal 2 (Framework for performance monitoring): Support reusing the framework for performance monitoring configuration for AI/ML based BM, where the NW configures one CSI report configuration for inference and another linked configuration for monitoring.
- Proposal 3 (Performance metric measurement): Support the CQI offset between the CQI for the predicted CSI and the CQI for the ground-truth CSI as the metric for performance monitoring.
- Proposal 4 (Performance metric measurement): Support the NW configuring separate CSI-RS resource sets for inference and monitoring, allowing CSI-RS to be periodic, semi-persistent, or aperiodic.
- Proposal 5 (Report mechanism): Support NW-triggered performance monitoring results report, while deprioritizing event-triggered performance results reports.
- Proposal 6 (Report mechanism): Support reporting the performance monitoring results by UCI to ensure low latency comparable to CSI reports.