R1-2410564
discussion
AI/ML for CSI prediction
From Mavenir
Summary
Mavenir presents evaluation results comparing AI/ML-based CSI prediction using LSTM networks against Kalman filter baselines, demonstrating superior performance across different UE speeds. The document contains 6 proposals and 3 observations focusing on CSI prediction implementation, performance monitoring, and fallback mechanisms for AI/ML in NR air interface.
Position
Mavenir advocates FOR AI/ML-based CSI prediction using LSTM networks with UE-side implementation, emphasizing the need for enhanced CSI-RS configurations and proper fallback mechanisms. They push FOR leveraging uplink measurement reference signals in TDD mode to reduce CSI-RS overhead while maintaining prediction accuracy, and advocate FOR correlation-based performance monitoring using SGCS metrics between legacy and AI-predicted CSI.
Key proposals
- Proposal 1 (Sec 2.3): Channel measurement error needs to be considered when evaluating and improving CSI prediction models
- Proposal 2 (Sec 2.3): Enhanced CSI-RS configurations are critical for AI/ML model training because models need accurate historical CSI data to learn the changing patterns of the channel
- Proposal 3 (Sec 2.3): Further discuss in TDD mode, how the uplink measurement reference signal can be utilized to reduce the overhead of the CSI-RS, while improving the downlink CSI prediction accuracy
- Proposal 4 (Sec 2.4): Regarding CSI prediction with UE-sided model, for NW side performance monitoring, using an existing CSI feedback scheme as the reference can be considered
- Proposal 5 (Sec 2.4): The fallback to legacy CSI prediction mechanism should be set to achieve balance of performance and complexity
- Proposal 6 (Sec 2.4): The correlation metrics(SGCS) between the legacy CSI and the predicted CSI can be used to monitor the performance of AI based CSI prediction