R1-2500203
discussion
Discussion on AI/ML-based CSI prediction
From CATT
CATT's prior position on
9.1.3
at
RAN1#119
· AI-synthesized, paraphrased
verify sources →
Advocates for relaxing consistency requirements for network-side antenna configurations like tilt angles and TXRU mapping, demonstrating through simulations that they have negligible impact. Pushes for performance monitoring-based methods to handle other consistency factors, such as interference distributions, which cannot be addressed through associated IDs. Positions against overly restrictive consistency requirements that would unnecessarily complicate CSI prediction implementations.
Summary
CATT presents simulation results demonstrating that antenna tilt angles and TXRU mappings have negligible impact on UE-sided CSI prediction performance, concluding that strict consistency between training and inference for these parameters is unnecessary. The document proposes introducing a new processing unit type for AI/ML inference, distinguishing AI/ML reports from legacy ones, and deprioritizing Type 2 performance monitoring due to overhead and accuracy concerns, while supporting SGCS as the metric for Type 1 and Type 3 monitoring.
Position
CATT concludes that consistency between training and inference regarding antenna down tilt and TXRU mapping is not required, citing simulation results showing negligible performance impact on SGCS. They propose introducing a new processing unit type or enhanced CPU for AI/ML-based CSI processing, which is separately counted from the legacy CPU pool but shared among CSI-related AI/ML functionalities. CATT supports distinguishing AI/ML CSI reports from legacy reports via a new report quantity or identifier. For performance monitoring, they prefer SGCS as the metric for Type 1 and Type 3 monitoring and explicitly deprioritize Type 2 monitoring due to high overhead and quantization errors. They propose reusing the CSI framework for monitoring configuration, allowing either reuse of inference resources or dedicated monitoring resources.
Key proposals
- Proposal (Sec 2.2): Conclude that there is no need to ensure consistency between training and inference regarding antenna down tilt and TXRU mapping configuration at network side for CSI prediction.
- Proposal (Sec 3.1): Support distinguishing AI/ML-based CSI report from legacy CSI report by introducing a new report quantity or an identifier in the CSI report configuration.
- Proposal (Sec 3.2): Introduce a new processing unit type or enhanced CPU dedicated to AI/ML-based CSI processing, separately counted from the legacy CPU pool and shared among CSI-related AI/ML features.
- Proposal (Sec 3.2): Require AI/ML-based CSI reporting to meet the requirements of both legacy CPU and the new processing unit type or enhanced CPU.
- Proposal (Sec 3.2): UE should report information necessary to determine the number of occupied new processing unit type or enhanced CPU.
- Proposal (Sec 3.2): Enhance CPU occupation rules to specify the symbols during which the new processing unit type or enhanced CPU is occupied for UE-sided CSI prediction.
- Proposal (Sec 3.2): Determine the priority of CSI reports for different functionalities, data collection, model inference, and performance monitoring based on report quantity and other configuration parameters.
- Proposal (Sec 4.1): Support SGCS (similarity between reported inference result and ground-truth eigenvector) as the performance metric for UE-side CSI prediction.
- Proposal (Sec 4.1): For Type 3 performance monitoring, discuss reporting options including metrics for each layer/slot interval, specific layers/intervals, or averages over multiple intervals.
- Proposal (Sec 4.1): Support both network-configured (periodic, semi-periodic, aperiodic) and event-triggered reporting for Type 1 and Type 3 performance monitoring, with events defined by reference to beam failure events.
- Proposal (Sec 4.1): Deprioritize Type 2 performance monitoring for UE-side CSI prediction due to high reporting overhead of channel matrix reporting and accuracy impacts of codebook-based reporting.
- Proposal (Sec 4.2): Reuse the CSI framework for configuring monitoring reports for Type 1 and Type 3 performance monitoring, linking the inference report ID within the monitoring report configuration.
- Proposal (Sec 4.2): Support options for monitoring resource configuration: reusing resources from the inference report or configuring dedicated resources for monitoring.