Wireless AI Model Retraining Control for CSI and Beam Management
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing mobile communication systems face challenges in effectively managing and utilizing AI/ML models for wireless communication, particularly in managing model changes and re-training processes, which can lead to inefficiencies and suboptimal performance.
Innovation Solution
A communication control method and user equipment that include a model management entity for tracking model change information and a controller for managing re-training of AI/ML models, enabling efficient model training and inference processes, such as CSI feedback enhancement, beam management, and positioning accuracy enhancement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If re-training is performed on a trained first AI/ML model to improve performance, then model accuracy is improved, but overhead and power consumption increase
Solution Approach 1:
The patent changes the state parameter of the AI/ML model by performing re-training operations. The model management entity controls whether re-training is performed on the trained first AI/ML model, allowing the system to adapt model parameters (weights and biases) based on new data while managing the trade-off between accuracy improvement and power consumption through selective re-training
2Measurement precision
If re-training is performed on a trained first AI/ML model to improve performance, then model accuracy is improved, but system overhead increases
Solution Approach 1:
The patent introduces a model management entity as an intermediary between the AI/ML model and the re-training process. This intermediary manages model change information and controls whether re-training is performed, thereby reducing system overhead by centralizing model management functions and avoiding direct complex interactions between multiple components during re-training operations
3Reliability
If model change information is tracked to manage re-training, then model management reliability is improved, but device complexity increases
Solution Approach 1:
The patent creates and manages model change information as a copy or metadata representation of the AI/ML model state. The model management entity tracks whether re-training has been performed by maintaining this information copy, which allows reliable model management without requiring complex real-time monitoring of the actual model parameters, thereby reducing the complexity overhead
Data Source
AI summary
The present disclosure relates to a communication control method in a mobile communication system. The communication control method includes managing, by a model management entity, model change information indicating whether model training has been performed on a trained first AI/ML model.