R1-2410201
discussion
Discussion on AI/ML for CSI compression
From KAIST
Summary
KAIST proposes enhancements to AI/ML-based CSI compression for temporal domain aspects, specifically addressing non-ideal UCI feedback scenarios. The document contains 2 main proposals focusing on incorporating additional information beyond UCI loss probability for better CSI reconstruction and historical information management.
Position
KAIST advocates FOR incorporating multiple error indicators (ACK/NACK probability, data error probability) beyond just UCI loss probability when managing historical CSI information in temporal domain AI/ML compression. They push FOR more sophisticated error evaluation mechanisms that consider channel quality degradation even when UCI is successfully received, arguing that reconstruction quality may still be inappropriate for current channel conditions.
Key proposals
- Proposal 1 (Sec 3): For temporal domain aspect Case 2, consider the UCI loss probability along with other information (e.g. ACK/NACK probability)
- Proposal 2 (Sec 3): On non-ideal UCI feedback for temporal domain CSI compression, consider other information to reset historical CSI information in both UE and NW or to retrain the NW and/or UE model, and consider how to reuse historical CSI information by evaluating other information (e.g. data error probability) in both UE and NW