AI Channel Estimation for Diverse DMRS Configurations

The patent segments the channel estimation task by creating multiple AI models, each specialized for specific DMRS configurations (e.g., different PRB bundling sizes, PDSCH mapping types, DMRS counts). This segmentation allows each model to focus on extracting features from particular configuration types, thereby improving estimation accuracy without requiring a single overly complex universal model.

The patent applies local quality by tailoring AI models to specific local conditions (DMRS configurations). Each AI model is optimized for particular configuration parameters, ensuring that the model structure and parameters are locally adapted to the characteristics of each configuration type, which improves feature extraction capability for that specific configuration.

The patent implements a dynamic model selection mechanism that adapts the AI model based on the detected DMRS configuration. The terminal device dynamically selects the appropriate AI model according to the specific configuration parameters (PRB bundling size, PDSCH mapping type, DMRS count, CDM groups), ensuring that the system complexity is only as high as necessary for the current configuration rather than maintaining maximum complexity for all possible configurations simultaneously.

The patent changes the parameter of model selection based on configuration parameters. By adjusting which AI model is active based on DMRS configuration parameters, the system optimizes the balance between accuracy and complexity. The terminal device receives configuration information and selects the corresponding AI model, thereby adapting the system complexity to match the required precision for each specific configuration scenario.