NR CSI Feedback Compression Using Shared Autoencoder Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In 3GPP New Radio (NR), accurately describing high-dimensional channel conditions for CSI feedback requires a large number of bits, and existing methods struggle to enhance accuracy in CSI reporting.
Innovation Solution
Implementing autoencoders at both UE and base station to compress CSI feedback, with initial configuration of supported autoencoder structures and weights/parameters shared via MasterInformationBlock or SystemInformationBlock, followed by iterative updates based on error measurements to refine the autoencoder components.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-dimensional channel condition parameters are transmitted to accurately describe channel conditions, then measurement precision is improved, but the quantity of information increases requiring more bits for transmission
Solution Approach 1:
The patent segments the CSI feedback process into two parts: (1) transmitting a compressed low-dimensional representation of channel conditions using autoencoder, and (2) transmitting separate autoencoder configuration parameters (weights and structures) that enable the base station to reconstruct the original high-dimensional channel state. This segmentation allows accurate channel description without directly transmitting all high-dimensional parameters.
Solution Approach 2:
The patent introduces autoencoder models as an intermediary mechanism between UE and base station. The autoencoder compresses high-dimensional channel states into low-dimensional representations for transmission, while the base station uses the same autoencoder model (shared via configuration parameters) to decompress and reconstruct the original channel state information, maintaining accuracy without direct high-dimensional transmission.
2Measurement precision
If autoencoder structures and weights are shared between UE and base station to improve CSI reporting accuracy, then measurement precision is improved, but device complexity increases due to multiple autoencoder components
Solution Approach 1:
The patent makes the autoencoder model universal by configuring both UE and base station with the same autoencoder structure and weights through standardization parameters. This single autoencoder model serves multiple functions: compression at UE, decomposition at base station, and enables accurate CSI reporting without requiring different models for different devices.
Solution Approach 2:
The patent manages complexity by parameterizing the autoencoder through standardized configuration parameters (weights and structures) that can be dynamically adjusted. Instead of implementing complex fixed autoencoders, the system uses parameter changes to adapt the same autoencoder structure to different channel conditions and requirements, simplifying implementation while maintaining accuracy.
Data Source
AI summary
The present disclosure includes a method for managing channel state information feedback compression in user equipment. The method includes: receiving at least one reference signal; estimating a channel condition based on the at least one reference signal; reducing, using a first autoencoder encoder in user equipment (UE), a dimension of the estimated channel condition such that a low dimension channel condition is generated; calculating, using a second autoencoder encoder and an autoencoder decoder in the UE, an error measurement based on a difference between UE channel state input and UE channel state output; and transmitting at least one of a channel state information bit stream or at least one message, wherein the at least one of the channel state information bit stream or the at least one message is based on at least one of the low dimension channel condition or a value indicative of the error measurement.