In the dynamic world of automated driving functions, securing #AI-powered sensor data processing is crucial. With our new project “Praxis,” we are launching an innovative research initiative dedicated to precisely this topic. The goal is to ensure the safety and efficiency of automated driving functions through advanced testing methods and realistic simulation environments.
Project goals
The “Praxis” project pursues several key goals:
- Development of xAI-based testing methods: We are relying on explainable AI (xAI) to increase the transparency and traceability of our testing methods.
- Establishment of a realistic simulation environment: By creating an environment that simulates real conditions, we can test the performance of our algorithms under controlled, but realistic scenarios.
- Securing ML algorithms directly on target hardware: We integrate our tests directly into the target hardware to ensure the practical suitability of our solutions.
- Transferring ML methods into classic software and hardware tests: We combine modern machine learning methods with proven testing approaches to enable comprehensive validations.
- Prototype development of an “Over-the-Air” camera test bench: This test bench enables efficient validation and benchmarking of camera systems used in automated vehicles.
Our approach: Camera-in-the-Loop Testing
A central element of our project is the novel test bench, which tests real cameras under controlled conditions with simulated scenarios. The heart of this approach is a modified projector, which transmits HDR content from the CARLA simulation environment optically to the camera – synchronized to the millisecond.
This innovative technique creates combined real and synthetic image data, enabling robust validation of perception algorithms. The acquired data, consisting of RGB and depth information, allows for a reliable evaluation of the AI under realistic, yet fully reproducible conditions. With “Praxis,” we are thus making a valuable contribution to the safe and scalable development of automated driving functions.
That’s why we’re involved!
With the research project “Praxis,” we are building a technological advantage in the area of sensor performance validation. Our goal is to further refine and demonstrate our expertise in perception algorithms. We aim to achieve this competitive advantage with minimal financial investment for the next phase of growth.
We are convinced that “Praxis” not only increases the safety of automated driving functions but also represents a significant step towards responsible and innovative mobility. Stay tuned for further updates on our project and the exciting developments we expect in the coming months!