R1-2407798
discussion
Discussion on specification support for AI CSI prediction
From ZTE
Summary
ZTE proposes a methodical approach to AI CSI prediction consistency issues, advocating to first identify potential additional conditions before developing detailed solutions. The document contains 2 proposals focused on leveraging existing AI beam prediction mechanisms as a starting point.
Position
ZTE advocates FOR a systematic approach that prioritizes identifying additional conditions before rushing into detailed consistency solutions, and FOR reusing proven AI beam prediction mechanisms. They are AGAINST introducing new associated IDs for scenarios/carrier frequency, arguing these have limited performance impact (less than 2% degradation) and can be handled through existing configurations or UE implementation.
Key proposals
- Proposal 1 (Sec Necessity): RAN1 first identifies the potential additional conditions for AI CSI prediction before discussing the detailed solution for addressing consistency of training/inference
- Proposal 2 (Sec Necessity): If it is justified to address the potential consistency issue for AI CSI prediction, take the same mechanism of associated ID agreed for AI beam predication as starting point for AI CSI prediction
- Proposal 1 (Sec Conclusion): RAN1 first identifies the potential additional conditions for AI CSI prediction before discussing the detailed solution for addressing consistency of training/inference
- Proposal 2 (Sec Conclusion): If it is justified to address the potential consistency issue for AI CSI prediction, take the same mechanism of associated ID agreed for AI beam predication as starting point for AI CSI prediction