Wireless Node Channel Feedback with Trained RS Parameter Groups
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional PMI feedback methods in wireless communication systems with multiple antennas result in significant redundant overhead, necessitating a more efficient method for determining parameter groups used in channel information generation.
Innovation Solution
A method involving the use of a first and second information block to indicate a RS resource set and an identifier, where a code generates channel information based on trained parameter groups, allowing flexible adjustment and improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional PMI feedback method is used, then channel information can be obtained, but redundant overhead increases significantly
Solution Approach 1:
The patent extracts only the essential parameters needed for channel information generation by using AI/ML models to process and compress CSI-RS resource indicators, rank indicators, and precoding matrix indicators. This extraction approach removes redundant information while preserving critical channel characteristics, thereby reducing feedback overhead without sacrificing information accuracy.
Solution Approach 2:
The patent changes the parameter representation from traditional detailed PMI feedback to compressed AI/ML-generated parameters. By transforming the feedback mechanism to use trained models that output optimized parameters, the system achieves more efficient parameter transmission that maintains channel information quality while reducing the quantity of transmitted data.
2Quantity of substance
If AI/ML-based CSI compression is implemented, then feedback overhead is reduced, but implementation complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training AI/ML models with comprehensive channel information datasets before actual communication occurs. This preliminary training phase enables the models to automatically learn optimal parameter transformations, shifting the complexity from runtime computation to offline training, thereby simplifying the operational implementation while maintaining compression effectiveness.
Solution Approach 2:
The patent introduces AI/ML models as intermediary components between the channel measurement phase and the feedback transmission phase. These models act as mediators that automatically process raw channel measurements and generate compressed parameters, abstracting the complexity of channel characterization from the feedback mechanism and enabling more efficient communication with manageable implementation.
3Productivity
If more antennas are used, then transmission capacity increases, but PMI feedback redundancy increases
Solution Approach 1:
The patent replaces the traditional mechanical approach of explicitly feedbacking all PMI parameters for each antenna with an AI/ML-based information processing system. This substitution uses trained neural networks to compress and represent channel information more efficiently, enabling the system to handle increased antenna counts without linearly increasing feedback overhead, thus maintaining scalability with transmission capacity.
Data Source
Figure 1~3
Figure 4~5
Figure 6A~7C
AI summary
Disclosed in the present application are a method used in nodes for wireless communication, and an apparatus. The method comprises: a first node receiving a first information block, the first information block being used for determining a first RS resource set, and the first RS resource set comprising one or more RS resources; receiving a second information block, the second information block indicating at least one identifier; and sending first channel information. A first code is used for generating the first channel information, and a measurement for at least one RS resource in the first RS resource set is used for generating an input of the first code; the at least one identifier is used for determining a parameter group used by the first code, the parameter group used by the first code being one parameter group amongst J parameter groups, J being a positive integer greater than 1; and at least one parameter group amongst the J parameter groups is obtained by training.