Synthetic Sensor Data Generation Using Depth Maps for ADAS Testing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for testing autonomous systems, such as advanced driver-assistance systems (ADAS) and autonomous driving systems, face challenges in accurately simulating sensor data due to variations in sensor position and physical characteristics, leading to high costs and inefficiencies in data capture and processing.
Innovation Solution
A system and method using machine learning models to compute depth maps from real signals captured by multiple sensors, applying point of view and physical characteristic transformations to create synthetic data that simulates signals from a target sensor, reducing distortion and improving accuracy for testing and validation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real sensor data is captured from multiple positions and configurations to train and test autonomous systems, then the accuracy and reliability of testing improve, but the cost and complexity of data capture increase significantly
Solution Approach 1:
The patent creates synthetic sensor data by rendering virtual scenes that replicate the appearance and characteristics of real sensor data. Instead of physically capturing data from multiple sensor positions and configurations, the system generates synthetic copies that mimic what those sensors would capture, thereby maintaining testing reliability while eliminating the complexity of physical data capture from multiple configurations
Solution Approach 2:
The system varies parameters in the virtual scene rendering process to simulate different sensor positions, orientations, and physical characteristics. By changing rendering parameters rather than physical sensor configurations, the system achieves diverse training data without the complexity of physically reconfiguring multiple sensors
2Productivity
If synthetic data is generated without accurate depth information, then the generation process is simpler and faster, but the quality and usefulness of the synthetic data for training autonomous systems deteriorates
Solution Approach 1:
The patent introduces depth maps as an intermediary element in the synthetic data generation process. These depth maps serve as intermediate representations that capture geometric information about the virtual scene, enabling the generation of high-quality synthetic sensor data with accurate depth information while maintaining efficient rendering processes. The depth maps act as a bridge between simple scene geometry and complex realistic sensor output
Data Source
AI summary
A system for creating synthetic data for testing an autonomous system, comprising at least one hardware processor adapted to execute a code for: using a machine learning model to compute a plurality of depth maps based on a plurality of real signals captured simultaneously from a common physical scene, each of the plurality of real signals are captured by one of a plurality of sensors, each of the plurality of computed depth maps qualifies one of the plurality of real signals; applying a point of view transformation to the plurality of real signals and the plurality of depth maps, to produce synthetic data simulating a possible signal captured from the common physical scene by a target sensor in an identified position relative to the plurality of sensors; and providing the synthetic data to at least one testing engine to test an autonomous system comprising the target sensor.