Terminal Machine Learning Channel Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In next-generation mobile communication systems, the implementation of machine learning models for high-accuracy channel estimation and efficient resource use in radio communication is hindered by the lack of clear definitions regarding which entity implements and trains the ML model, leading to suboptimal communication throughput and quality.
Innovation Solution
A terminal equipped with a receiving section for channel state information reference signals (CSI-RS) and a control section that utilizes a machine learning model trained on CSI-RS configurations to perform prediction, enabling high-accuracy channel estimation and efficient resource allocation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If machine learning models are implemented for channel estimation without clear entity definitions, then communication quality may improve, but system complexity and implementation difficulty increase
Solution Approach 1:
The patent segments the ML model implementation into distinct entities: the base station trains the ML model using historical data, while the terminal executes the trained model for real-time channel estimation. This segmentation of training and execution functions across different network entities reduces individual entity complexity while maintaining high estimation accuracy.
Solution Approach 2:
The patent introduces a network side server as an intermediary that facilitates ML model training and management. The server collects historical data from base stations, trains the ML models, and distributes them to terminals, thereby simplifying the complexity of model training and execution for individual network entities.
2Productivity
If intensive studies on ML model implementation are conducted, then communication throughput and quality improve, but time required for deployment increases
Solution Approach 1:
The patent performs preliminary training of ML models on historical data collected during network operation. The base stations and network side servers prepare and train models in advance using accumulated historical information, enabling the models to be directly deployed for real-time channel estimation without requiring extensive real-time training, thus reducing deployment time while maintaining high throughput.
3Measurement precision
If ML models are trained using extensive historical data, then estimation accuracy improves, but data processing requirements and computational resources increase
Solution Approach 1:
The patent segments the data processing workload by having base stations perform local data collection and preliminary processing, while the network side server handles centralized model training using aggregated historical data. This segmentation allows accurate models to be trained using comprehensive historical data without requiring every terminal to process the entire dataset, thereby reducing individual computational resource consumption.
Data Source
AI summary
A terminal according to one aspect of the present disclosure includes a receiving section that receives a channel state information reference signal (CSI-RS), and a control section that performs, by using a machine learning model trained based on first configuration of the CSI-RS, control of performing prediction based on second configuration of the CSI-RS. According to one aspect of the present disclosure, high-accuracy channel estimation, high-efficiency use of resources, and the like can be implemented.