Neural Network Demapping for Multi-User MIMO Resource Grids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wireless communication systems face challenges in efficiently demapping received transmissions, particularly in multi-user MIMO systems, where existing demodulation methods struggle with accuracy and adaptability to varying numbers of spatial streams and user apparatuses.
Innovation Solution
Employing a neural network-based demapper model that processes two-dimensional resource grids using trained models per spatial stream, incorporating channel estimation and equalization, and utilizing two-dimensional neural networks to generate log-likelihood ratios for improved demapping accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional demodulation methods are used in multi-user MIMO systems, then device complexity is reduced, but demapping accuracy deteriorates due to inability to adapt to varying numbers of spatial streams and user apparatuses
Solution Approach 1:
The patent implements dynamic adaptability by training separate neural network demapper models for different configurations of spatial streams and user apparatuses. The system selects or switches between pre-trained models based on the current transmission configuration, enabling the demapper to dynamically adapt to varying system conditions without requiring real-time retraining or complex on-the-fly adjustments.
Solution Approach 2:
The patent employs pre-trained neural network demapper models for various spatial stream and user apparatus configurations. These models are trained in advance during an offline phase, and the appropriate pre-trained model is selected based on the current system configuration. This preliminary action eliminates the need for complex real-time adaptation mechanisms during actual demapping operations.
2Measurement precision
If neural network-based demappers are trained for each spatial stream configuration, then demapping accuracy is improved, but device complexity increases due to multiple trained models
Solution Approach 1:
The patent creates universal demapper models that can handle multiple spatial stream configurations and user apparatus combinations. Each trained model is designed to be multi-functional, capable of processing different numbers of spatial streams and serving various user apparatus configurations. This universality reduces the total number of specialized models needed while maintaining high demapping accuracy across different scenarios.
Solution Approach 2:
The patent employs multiple copies of demapper models, where each copy is trained for a specific configuration but shares the same underlying architecture and training methodology. These copied models can be efficiently stored and selected based on current system conditions, reducing the computational overhead compared to developing entirely different demapper architectures for each configuration.
3Adaptability or versatility
If existing demodulation methods are used, then ease of operation is maintained, but adaptability to varying numbers of spatial streams and user apparatuses deteriorates
Solution Approach 1:
The patent performs preliminary training of neural network demapper models for various spatial stream and user apparatus configurations during an offline phase. This preliminary action prepares multiple pre-configured models that can be directly deployed without requiring complex real-time adjustments, thereby maintaining operational simplicity while achieving high adaptability to different system configurations.
Solution Approach 2:
The patent implements dynamic model selection or switching mechanisms that automatically choose the appropriate pre-trained demapper model based on the current number of spatial streams and user apparatuses. This dynamic adaptation occurs through simple model selection rather than complex reconfiguration, preserving ease of operation while achieving versatility across different transmission scenarios.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
To provide demapping at a receiving side, a trained model for a demapper is used to output log-likelihood ratios of received signals representing data in a multi-user transmission. Inputs for the trained model for the demapper comprise a resource grid of equalized received signals.