ML Channel Feedback Interoperability for Accurate CSI Reporting

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing wireless communication systems face challenges in efficiently reporting channel conditions due to high resource consumption and inadequate accuracy using codebooks, and collaboration between apparatus and vendors is difficult without standardization for model interoperability.

Innovation Solution

Implement machine learning models for channel information feedback, enabling interoperability through model specification, capability indication, and data transfer protocols, allowing for reduced resource usage and improved accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If codebooks are used to reduce feedback data transmission, then resource consumption is reduced, but feedback accuracy deteriorates

Engineering Contradiction:
Improvefeedback data sizeVSAvoidchannel condition feedback accuracy
Core Design Contradiction:
Quantity of substanceVSMeasurement precision

Solution Approach 1:

The patent changes the fundamental parameters of feedback representation from discrete codebook indices to continuous compressed channel information using machine learning models. The encoder model compresses channel state information into a compact representation that captures essential channel characteristics, achieving both reduced data size and maintained accuracy through intelligent parameter transformation rather than simple quantization

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces the mechanical/codebook-based feedback system with an intelligent machine learning-based compression system. Instead of relying on predefined discrete codebooks, the system uses neural network encoders and decoders to transform channel information into compressed representations, enabling more efficient and accurate feedback transmission

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If machine learning models are used to compress channel information, then resource consumption is reduced, but model interoperability and collaboration complexity increases

Engineering Contradiction:
Improvefeedback data sizeVSAvoidmodel interoperability complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent introduces standardized model information structures and capability indication mechanisms as intermediaries between different devices and vendors. These standardized interfaces enable machine learning models to interoperable communication without requiring direct collaboration between all components, reducing the complexity of model integration while maintaining compression benefits

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent creates universal model information formats and capability indication protocols that can be used across different device types and vendor implementations. This universality allows the same compression mechanism to work in various communication scenarios without requiring custom integration for each case, simplifying deployment

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4645768A1Model interoperability and collaboration for machine learning based reporting of communication channel information
Publication Date: 2025.11.05 SAMSUNG ELECTRONICS CO LTD
  • EP4645768A1 patent drawingFigure 1~2
  • EP4645768A1 patent drawingFigure 3
  • EP4645768A1 patent drawingFigure 4

AI summary

An apparatus (3101) may include a receiver configured to receive a reference signal (3117) using a channel (3115), a transmitter configured to transmit a representation (3107) relating to the channel (3115), and a processing circuit configured to determine channel information based on the reference signal, generate, using a model, the representation based on the channel information, and transfer, using the receiver or the transmitter, at least one of the following: (a) model information (3173A) to specify the model, (b) capability information (3173B) for the apparatus relating to the model, and (c) data set information (3173C) to specify a data set for the model. The model information (3173A) may include an identifier for the model. The model information (3173A) may include structure information for the model. The model information (3173A) may include information about a type of input for the model. The model information (3173A) may include information about a format of input for the model. The model information (3173A) may include mapping information for mapping channel information to an input of the model. The mapping information may include information for a first subband, and information for a second subband.