Radar Scatter Plot Transformation for Realistic Training Data

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The scarcity and difficulty of obtaining high-quality training data for object recognition from radar signals, due to the specialized knowledge required and sensor-specific distortions, result in residual uncertainty when using synthetic data for real-world applications.

Innovation Solution

A method to generate more realistic simulated radar data by imposing the influence of physical properties from real-world measurements onto simulated data using a Generative Adversarial Network (GAN), transforming point clouds through density distributions to account for sensor-specific distortions and noise.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Quantity of substance

If synthetic radar data is used for training, then data availability is improved, but realism and sensor-specific characteristics are worsened

Engineering Contradiction:
Improvetraining data availabilityVSAvoidrealism of training data
Core Design Contradiction:
Quantity of substanceVSReliability

Solution Approach 1:

The patent creates synthetic radar data by generating point clouds that copy the essential structure and characteristics of real radar measurements. By simulating radar reflections, noise patterns, and ghosting effects through computational models, the system produces training data that replicates real sensor behavior without requiring physical measurement campaigns.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent introduces an intermediate processing layer that transforms synthetic point cloud data into a format that better matches real radar measurements. This intermediary transformation applies sensor-specific distortion models, noise characteristics, and signal processing effects to bridge the gap between idealized simulations and actual sensor outputs.

Inventive Principle:
Principle #24Intermediary (Mediator)

2Measurement precision

If manual annotation of radar data is performed, then training data quality is improved, but time and expertise requirements are worsened

Engineering Contradiction:
Improvetraining data qualityVSAvoidannotation time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent implements self-service annotation through automated object detection and labeling algorithms that process radar point clouds without human intervention. The system automatically identifies objects, classifies them, and generates ground truth labels by analyzing the spatial distribution and reflection characteristics of points in the synthetic data.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent performs preliminary annotation during the data generation phase by embedding known object positions and characteristics into the synthetic scene models before radar signal simulation. This allows training labels to be automatically associated with generated data without requiring post-processing annotation steps.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If sensor-specific distortions are included in simulation, then training data realism is improved, but simulation complexity is worsened

Engineering Contradiction:
Improvetraining data realismVSAvoidsimulation complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent segments the complex simulation process into distinct modular components: point cloud generation, radar signal simulation, noise addition, and distortion application. Each module handles a specific aspect of sensor characteristics independently, allowing for systematic implementation and reducing overall system complexity through functional decomposition.

Inventive Principle:
Principle #1Segmentation

Data Source

PatentEP3818467B1More realistic simulation of physical measurement data
Publication Date: 2026.02.11 ROBERT BOSCH GMBH
  • EP3818467B1 patent drawingFigure 1
  • EP3818467B1 patent drawingFigure 2a~2d
  • EP3818467B1 patent drawingFigure 3a~3d

AI summary

A method (100) for impressing the influence of a physical property (Ia), which is shared by the measurement data (1) obtained by physical measurement and contained in at least one learning scatter plot PL in a domain B, onto simulated measurement data (2) in a simulation scatter plot PA in a domain A, the measurement data (1, 2) in the scatter plots PL and PA in each case representing coordinates, and the method comprising the following steps: • the simulation scatter plot PA is converted into a density distribution pA in the domain A (110); • the density distribution pA is converted by a transformation into a density distribution pB in the domain B (120), wherein this transformation is such that in domain B it is indistinguishable whether a given density distribution p was obtained directly in the domain B as a density distribution pB <sb />of a learning scatter plot PL or as transformation pe of a density distribution pA; • a result scatter plot PB, which is statistically consistent with the density distribution pe is produced in the domain B (130); the result scatter plot PB is assessed as result of the impressing of the influence of the desired property (1a) on the simulated measurement data (2) in the simulation scatter plot PA (140). The invention also relates to: a training method (200); a data set obtained by means of a method (100); a trained AI module and a corresponding data set; a method (300) for identifying objects (5a) and situations (5b); and an associated computer program.