MIMO CSI Feedback Compression via Online Encoder-Decoder Training

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current wireless communication systems face significant strain and resource consumption due to the high overhead of raw channel state information (CSI) feedback, especially in multiple-input and multiple-output (MIMO) technology, which limits system performance and increases delay.

Innovation Solution

Implementing a method for compressing CSI using an encoder-decoder model pair, where the user equipment (UE) and base station (BS) perform online training to generate updated models, allowing for efficient compression and decompression of CSI data, reducing the feedback overhead and complexity.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If raw CSI feedback is used in MIMO systems, then channel state information accuracy is improved, but communication resource consumption increases and system performance deteriorates

Engineering Contradiction:
Improvechannel state information accuracyVSAvoidcommunication resource consumption
Core Design Contradiction:
Measurement precisionVSLoss of energy

Solution Approach 1:

The patent extracts only the essential channel state information features using encoder-decoder models and neural network-based compression techniques, rather than transmitting complete raw CSI data. This selective extraction maintains measurement precision while significantly reducing feedback overhead and communication resource consumption.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent transforms CSI feedback from raw high-dimensional data to compressed low-dimensional representations by changing the parameter representation through online training of encoder-decoder models. This parameter transformation preserves critical channel state information while reducing the volume of feedback data, thereby lowering energy consumption.

Inventive Principle:
Principle #35Parameter changes

2Measurement precision

If raw CSI feedback is used in MIMO systems, then channel state information accuracy is improved, but feedback overhead increases

Engineering Contradiction:
Improvechannel state information accuracyVSAvoidfeedback overhead
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential channel state information features using encoder-decoder models and neural network-based compression techniques, rather than transmitting complete raw CSI data. This selective extraction maintains measurement precision while significantly reducing feedback overhead and communication resource consumption.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies dimensionality reduction by transforming high-dimensional raw CSI into low-dimensional compressed representations through online training of encoder-decoder models. This dimensional transformation preserves critical channel information while drastically reducing feedback overhead.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

3Measurement precision

If raw CSI feedback is used in MIMO systems, then channel state information accuracy is improved, but system complexity increases

Engineering Contradiction:
Improvechannel state information accuracyVSAvoidsystem complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent implements self-service through autonomous online training of encoder-decoder models at the user equipment. The system automatically adapts to changing channel conditions and updates compression parameters without requiring complex centralized coordination or manual configuration, thereby reducing overall system complexity while maintaining accuracy.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent introduces dynamic adaptation through online training mechanisms that continuously update encoder-decoder models based on current channel conditions. This dynamic approach replaces static complex processing with adaptive lightweight models, reducing system complexity while preserving measurement precision.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentUS20240380443A1Method and apparatus for multiple-input and multiple-output (MIMO) channel state information (CSI) feedback
Publication Date: 2024.11.14 MEDIATEK SINGAPORE PTE LTD
  • US20240380443A1 patent drawing
  • US20240380443A1 patent drawing
  • US20240380443A1 patent drawing

AI summary

This disclosure provides methods for channel state information (CSI) feedback. In one method, at a user equipment (UE), CSI data of a communication channel between the UE and a base station (BS) is collected. The UE is configured with an encoder model to compress CSI and the BS is configured with a decoder model to decompress CSI. Based on the collected CSI data, online training is performed at the UE on a previously trained encoder-decoder model pair including the encoder model and the decoder model to generate updated models for the encoder model and the decoder model, respectively. In another method, the CSI data is collected at the BS and the online training is performed at the BS based on the collected CSI data. In another method, the CSI data is collected at a server and the online training is performed at the server based on the collected CSI data