Multi-TRP AI/ML Model Management for Wireless Signal Transmission

The patent segments AI/ML model management by introducing separate MAC CE and DCI mechanisms for different management functions. MAC CE handles model configuration and activation/deactivation, while DCI handles real-time model switching and updating. This segmentation allows complex AI/ML management to be divided into manageable signaling layers, improving adaptability while controlling complexity.

The patent implements dynamic AI/ML model management where models can be activated, deactivated, updated, or switched based on real-time network conditions and UE capabilities. The DCI enables dynamic model switching between multiple TRPs, while MAC CE provides semi-static configuration. This dynamic approach enhances adaptability without requiring permanent complex configurations for all scenarios.

The patent creates a universal AI/ML model management framework that works across multiple TRPs using common MAC CE and DCI structures. The same signaling mechanisms (MAC CE for configuration, DCI for activation) are used universally for each TRP, providing consistent reliability without requiring TRP-specific complex configurations. This multi-functional approach handles model management uniformly across different TRPs.

The patent implements preliminary action by configuring AI/ML models in advance through MAC CE before actual transmission/reception operations. The network pre-configures multiple AI/ML models for different TRPs, and the UE stores these configurations ready for rapid activation via DCI. This preliminary configuration reduces real-time complexity while ensuring reliable model availability for each TRP when needed.

The patent applies preliminary action by pre-configuring multiple AI/ML models through MAC CE before they are needed for transmission. The network prepares model configurations in advance and stores them at the UE, so when switching between TRPs is required, the UE can rapidly activate pre-prepared models via DCI without time-consuming configuration processes. This reduces the time loss associated with frequent model updates and switching.

The patent implements local quality by maintaining different AI/ML models locally at the UE for different TRPs, each optimized for specific transmission conditions. Instead of centrally managing all model operations, the UE locally stores and rapidly switches between pre-configured models based on DCI indications. This local model storage and switching mechanism enables fast productivity improvements while minimizing centralized management time overhead.