ML-Based OFDM Resource Grid Mapping for Spectral Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OFDM-based wireless communication systems face challenges in efficiently utilizing resources due to tempo-spectral variations in channel conditions, leading to increased error rates, inter-subcarrier interference, and difficulties in link adaptation and scheduling.
Innovation Solution
Implementing a machine learning algorithm with trainable parameters, utilizing artificial neural networks at both the transmitter and receiver to spread modulated data symbols across the entire time-frequency resource grid, optimizing the transmission and reception processes to enhance performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional OFDM modulation is used, then the system is simple to implement, but spectral efficiency is limited due to tempo-spectral variations in channel conditions
Solution Approach 1:
The patent replaces traditional mechanical signal processing methods with machine learning algorithms. Specifically, neural networks are used to learn optimal resource grid mappings and signal transformations, substituting conventional fixed algorithms with adaptive learning-based approaches that improve spectral efficiency while managing complexity through automated training
Solution Approach 2:
The patent changes the parameters of the communication system by introducing trainable parameters in machine learning models. These parameters are optimized during training to adapt to tempo-spectral variations in channel conditions, allowing the system to dynamically adjust to changing spectral efficiency requirements without manual intervention
2Productivity
If resource allocation is optimized for specific channel conditions, then spectral efficiency improves, but the system becomes difficult to adapt to varying channel conditions
Solution Approach 1:
The patent introduces dynamic adaptability through machine learning models that can adjust their parameters based on varying channel conditions. The neural networks are trained to recognize and adapt to different tempo-spectral patterns, enabling the system to dynamically optimize spectral efficiency across varying link conditions rather than being fixed for specific scenarios
Solution Approach 2:
The patent implements feedback mechanisms where the machine learning models receive information about channel conditions and performance metrics, then adjust their resource allocation decisions accordingly. This closed-loop approach allows the system to continuously adapt to changing conditions while maintaining optimized spectral efficiency
3Reliability
If conventional OFDM demodulation is used, then the receiver is simple, but bit error rates increase under tempo-spectral channel variations
Solution Approach 1:
The patent replaces conventional fixed demodulation algorithms with machine learning-based demodulation models. These neural networks learn to inverse the transformations applied at the transmitter, providing more accurate signal recovery under tempo-spectral variations while the training process automates the complexity management
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed is a method comprising providing a first resource grid as input to a first machine learning algorithm, obtaining a second resource grid as output from the first machine learning algorithm, and transmitting a signal comprising the second resource grid by using orthogonal frequency-division multiplexing modulation.