Wireless Signal Training System Using Ray Tracing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The lack of suitable training data for machine learning systems to identify and classify modulation modes and encoding schemes of wireless signals, exacerbated by signal perturbations and distortions during propagation, hinders the development of reliable classification systems.
Innovation Solution
A training system that uses real-world measurements and digital terrain elevation data to predict channel perturbations, augmenting synthesized modulation data with hardware/software perturbations to generate robust training data for machine learning systems, such as neural networks, to simulate and classify wireless signals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real-world wireless signals are used for training machine learning systems, then the training data reflects actual propagation conditions, but the signals have undergone perturbation and distortion that cannot be readily reversed or sufficiently characterized to produce reliable truth data
Solution Approach 1:
The patent creates synthetic copies of wireless signals with known ground truth labels by simulating propagation through digital terrain elevation data and channel models. These synthetic signals replicate real-world distortion effects while maintaining known original characteristics, enabling supervised training without requiring reversal of actual signal degradation
Solution Approach 2:
The patent applies channel perturbations and distortions to synthetic signals before training, pre-characterizing the distortion effects through digital terrain analysis and ray tracing. This preliminary characterization of propagation effects allows the machine learning system to learn distortion patterns in advance, improving robustness when classifying actual perturbed signals
2Productivity
If machine learning systems are trained to identify and classify modulation modes and encoding schemes, then classification speed and accuracy improve, but adequate training data with proper labeling is scarce
Solution Approach 1:
The patent generates large quantities of labeled training data by creating synthetic wireless signals with known modulation modes and encoding schemes. These synthetic copies can be produced in unlimited quantities with automatic ground truth labeling, eliminating the scarcity of annotated training examples needed for effective machine learning
Solution Approach 2:
The patent varies multiple parameters in synthetic signal generation including modulation types, encoding schemes, terrain configurations, and channel conditions to create diverse training examples. This parameter variation generates a comprehensive dataset covering the full range of possible signal characteristics the system must classify
3Quantity of substance
If synthesized modulation data is used for training, then adequate training data can be generated, but the data lacks real-world channel perturbation characteristics
Solution Approach 1:
The patent applies real-world channel perturbation models to synthetic training signals before training, using digital terrain elevation data and ray tracing to pre-characterize propagation effects. This preliminary application of distortion models ensures training data incorporates authentic channel characteristics, improving adaptability to real-world conditions
Solution Approach 2:
The patent introduces channel simulation models and digital terrain data as intermediaries between synthetic signal generation and machine learning training. These intermediaries transform idealized synthetic signals into realistic training examples by applying propagation effects, bridging the gap between controlled synthesis and real-world complexity
Data Source
AI summary
A signal generator outputs a reference signal corresponding to at least one wireless signal according to the predefined signal encoding to a channel emulator processor. The channel emulator processor is programmed to use at least one synthesized channel parameter and the reference signal to produce and store a perturbed signal as data for training machine learning and artificial intelligence systems. The synthesized channel parameter is synthesized using a channel synthesizer processor programmed to: ingest map elevation data, reference a transmitter and a receiver to the map elevation data, and perform ray tracing of a representative signal between the transmitter and the receiver, while applying at least one predetermined perturbation property to synthesize at least one channel parameter.