Autoencoder Beamforming Feedback Compression in WLAN

Solution Overview

Problem

Current wireless communication systems, such as IEEE 802.11 WLAN, face high computational complexity and communication overhead in beamforming feedback, particularly in training information updates, which hinders system throughput and efficiency.

Innovation Solution

The implementation of an autoencoder-based method that compresses and decompresses steering matrices using neural network quantization, reducing the need for extensive communication overhead by transmitting only quantized coefficients and scale factors, and utilizing a decoder part to obtain uncompressed steering matrices for beamformed data transmission.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Device complexity

If conventional beamforming feedback methods are used, then beamforming can be performed, but computational complexity and communication overhead increase significantly

Engineering Contradiction:

Improvecomputational complexityVSAvoidsystem throughput

Core Design Contradiction:

Device complexityVSProductivity

Solution Approach 1:

The patent extracts only the essential information (steering matrix) from the complete channel state information, and further compresses it using autoencoder-based neural network quantization. This extraction approach sends only quantized coefficients and scale factors instead of complete channel matrices, reducing communication overhead and computational complexity while maintaining beamforming functionality.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the representation parameters of the steering matrix by using neural network quantization to transform the continuous steering matrix into a discrete set of quantized coefficients and scale factors. This parameter transformation reduces the precision requirements while maintaining adequate beamforming performance, thereby reducing both communication overhead and computational complexity.

Inventive Principle:

Principle #35Parameter changes

2Measurement precision

If complete steering matrix feedback is transmitted, then beamforming accuracy is maintained, but communication overhead increases

Engineering Contradiction:

Improvebeamforming accuracyVSAvoidcommunication overhead

Core Design Contradiction:

Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential information (steering matrix) from the complete channel state information, and further compresses it using autoencoder-based neural network quantization. This extraction approach sends only quantized coefficients and scale factors instead of complete channel matrices, reducing communication overhead while maintaining beamforming functionality.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent uses an autoencoder neural network to create a compressed representation (copy) of the steering matrix. The encoder generates quantized coefficients that serve as a compressed copy of the original steering matrix, which can be reconstructed at the receiver using the decoder. This copying mechanism reduces communication overhead while preserving the essential beamforming information.

Inventive Principle:

Principle #26Copying

3Reliability

If frequent training information updates are performed, then beamforming performance is improved, but computational complexity and overhead increase

Engineering Contradiction:

Improvebeamforming performanceVSAvoidtraining update overhead

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent extracts only the essential information (steering matrix) from the complete channel state information, and further compresses it using autoencoder-based neural network quantization. This extraction approach sends only quantized coefficients and scale factors instead of complete channel matrices, reducing communication overhead and computational complexity while maintaining beamforming functionality.

Inventive Principle:

Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the representation parameters of the steering matrix by using neural network quantization to transform the continuous steering matrix into a discrete set of quantized coefficients and scale factors. This parameter transformation reduces the precision requirements while maintaining adequate beamforming performance, thereby reducing both communication overhead and computational complexity.

Inventive Principle:

Principle #35Parameter changes

Data Source

PatentUS20240380466A1Method and apparatus for low complexity beamforming feedback in wireless local area networks

Publication Date: 2024.11.14 SAMSUNG ELECTRONICS CO LTD

AI summary

A computer-implemented method performed by a first electronic device for reducing a feedback overhead of beamforming in a wireless communication system, includes: transmitting a first data to a second electronic device; transmitting a data packet to the second electronic device; receiving a second data from the second electronic device; extracting a compressed steering matrix from the second data; obtain uncompressed steering matrix by using a decoder part of an autoencoder, based on the extracted compressed steering matrix; and transmitting, to the second electronic device, a third data via a radio signal beamformed based on the obtained uncompressed steering matrix.