Synthetic Wireless Channel Data Generation via Latent Space AI
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The limited availability of measured datasets for wireless communication channels hinders effective testing and simulation of new measurement and communication equipment, as existing methods rely on real-world data that is often scarce.
Innovation Solution
A system comprising two AI units, an interface unit, and an analyzer unit that converts measured wireless channel data into synthetic data by transferring it into latent space and back, allowing for the detection of correlations and generation of tailored synthetic data with specific characteristics for testing scenarios.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-world measured datasets are used for testing and simulation, then the data reflects actual wireless communication channel behavior, but the quantity of available data is limited and scarce
Solution Approach 1:
The patent creates synthetic copies of real-world measured datasets using neural networks. The system trains a neural network model on available measured data and then generates multiple synthetic datasets that replicate the statistical properties and characteristics of real wireless channel measurements, thereby multiplying the available test data without requiring additional physical measurements
Solution Approach 2:
The patent transforms the limited measured datasets into a different representation (latent space) and then reconstructs them with modified parameters. By changing the parameter space through neural network transformations, the system generates diverse synthetic datasets that maintain the essential characteristics of real measurements while providing quantity variations
2Quantity of substance
If more measured datasets are collected from real-world measurements, then the quantity of data increases, but the time and resources required for measurement campaigns increase
Solution Approach 1:
The patent performs preliminary training of the neural network model using a small initial set of measured datasets. Once trained, the model can rapidly generate synthetic datasets without requiring additional time-consuming field measurements, thus achieving data multiplication after the initial preliminary training phase
Solution Approach 2:
Instead of performing additional time-consuming real-world measurements, the system creates synthetic copies of the measured data through neural network generation, significantly reducing the time required to obtain additional test datasets
3Adaptability or versatility
If synthetic data is generated without analyzing correlations with latent space attributes, then the generation process is simpler, but the ability to generate tailored synthetic data with specific characteristics is reduced
Solution Approach 1:
The patent implements feedback by analyzing correlations between latent space attributes and synthetic data characteristics. The system uses this correlation information to guide the generation process, allowing users to specify desired characteristics and have the neural network adjust the generation parameters accordingly, creating a feedback loop that enables precise control over synthetic data properties
Solution Approach 2:
The patent introduces latent space analysis as an intermediary step between data generation and user control. By analyzing correlations in the latent space, the system creates a bridge that allows users to influence specific characteristics of the generated synthetic data through controlled modifications in the latent representation
Data Source
AI summary
The present disclosure relates to a system for generating synthetic wireless channel data. The system comprises: an interface unit for receiving measured wireless channel data; a computing device comprising a first AI unit and a second AI unit to be trained; wherein the first AI unit is configured to transfer the measured wireless channel data into latent space data, and wherein the second AI unit is configured to convert the latent space data into synthetic wireless channel data; and wherein the first AI unit and the second AI unit are trained such that the synthetic wireless channel data resembles the measured wireless channel data; and an analyzer unit which is configured to produce correlation data which represents a correlation between the measured and/or the synthetic wireless channel data and at least one attribute of the latent space data.