Radar Reflection Point Generation via Power Distribution Pattern Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar simulation models, such as black-box and white-box models, face challenges in real-time simulation due to the lack of detailed environment models and high computational complexity, and deep stochastic radar models (DSRM) are limited by the inability to directly utilize real radar reflection points for further development.
Innovation Solution
A method that converts real radar reflection points into power distribution pattern images, allowing for model training and subsequent generation of radar reflection points, which can adapt to arbitrary roadway topologies and scene configurations, using a deep neural network to learn robust features from spatial data and object lists.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If white-box models use ray-tracing for estimating electromagnetic path propagation, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent uses black-box models to copy the input-output behavior of complex white-box ray-tracing models without replicating their internal complexity. The black-box model learns to map radar scene inputs to power return field outputs by training on synthetic data generated from ray-tracing simulations, thereby achieving similar measurement precision with reduced device complexity.
Solution Approach 2:
The patent replaces the mechanical ray-tracing computation system with a data-driven black-box model system. Instead of using explicit electromagnetic propagation calculations, the system uses trained neural networks that have learned the underlying patterns from ray-tracing data, substituting complex physical computation with optimized mathematical approximation.
2Measurement precision
If white-box models use detailed environment models to capture radar-related effects, then measurement precision is improved, but productivity decreases
Solution Approach 1:
The patent performs preliminary action by pre-generating training datasets using ray-tracing simulations with detailed environment models before deploying the black-box model. The complex computations are done in advance to create labeled training data, allowing the subsequent real-time application to use the pre-trained model without repeating the heavy computations, thus achieving both precision and productivity.
Solution Approach 2:
The black-box model copies the input-output characteristics of detailed environment models through training on pre-generated synthetic data. This allows the system to achieve accurate radar phenomena capture without requiring the detailed environment models to be present during real-time operation, enabling faster processing.
3Productivity
If deep stochastic radar models use neural networks for radar data processing, then productivity is improved, but device complexity increases
Solution Approach 1:
The patent extracts the essential input-output mapping capability from complex neural network models, using only the necessary components for the specific radar simulation task. The black-box model architecture is simplified to focus on learning the power return field generation from scene descriptions, removing unnecessary complexity while maintaining real-time processing capability.
4Device complexity
If black-box models represent radar return in a stochastic manner, then device complexity is reduced, but measurement precision decreases
Solution Approach 1:
The black-box model copies the deterministic input-output relationship of more complex models by training on synthetic data generated from ray-tracing simulations. Although the model structure is simple and stochastic, it learns to reproduce the accurate power return field patterns from the training data, achieving measurement precision comparable to white-box models despite the simpler architecture.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for generating radar reflection points comprising the steps of: providing a plurality of predefined radar reflection points of at least one first object detected by a radar and at least one first scenario description describing a first environment related to the detected first object; converting the predefined radar reflection points into at least one first power distribution pattern image related to a distribution of a power returning from the detected first object; training a model based on the first power distribution pattern image and the first scenario description; providing at least one second scenario description describing a second environment related to a second object; generating at least one second power distribution pattern image related to a distribution of a power returning from the second object based on the trained model and the second scenario description; and sampling the second power distribution pattern image.