R1-2410336
discussion
Discussion on AI/ML for CSI prediction
From AT&T
Summary
AT&T discusses AI/ML for CSI prediction in NR Air Interface, focusing on consistency between training and inference for UE-sided models. The document contains 6 observations and 3 main proposals advocating for network-side additional conditions and associated ID mechanisms to improve localized model performance.
Position
AT&T strongly advocates FOR site/cell specific AI/ML models over generalized models for CSI prediction, arguing they provide superior performance. They push FOR network-side additional conditions and associated ID mechanisms to enable localized model training and ensure training/inference consistency. They are advocating AGAINST relying solely on generalized models that show significant performance degradation across diverse scenarios.
Key proposals
- Observation (Sec 2): In CSI prediction using UE sided model use case, at least signaling and procedures for data collection, CSI-RS configuration, and assistance information for categorizing data have been proposed by companies
- Observation (Sec 6): For CSI prediction using UE-sided model, performance monitoring requires specification impacts including definition/configuration of performance metrics, threshold criteria, and report mechanisms for Type 1, 2, and 3 monitoring
- Proposal 1 (Sec 7): NW can provide NW side additional conditions to assist the UE to categorize the data set for different cell/site/scenario/configuration
- Proposal 2 (Sec 7): NW side additional conditions can be used to ensure consistency between the training and inference of the UE sided AI/ML model for CSI prediction
- Proposal 3 (Sec 7): For consistency between training and inference for CSI prediction using UE-sided model, further study the following option based on associated ID with FFS on what can be assumed by UE with the same associated ID across training and inference