Identification of scenarios with potential risk
Automated vehicles record dozens of terabytes of sensor data per day. Most of this data is irrelevant for testing. We identify the 0.1% of scenarios with potential safety risks.
Augmentation of real sensor data
We augment real sensor data with virtual objects to create safety-critical corner cases which are too rare to record in real traffic. We achieve a higher degree of realism than simulation while offering the same flexibility.
Safety analysis of your perception system
With our tools and metrics, we evaluate the safety of your system on challenging corner cases. We identify blind spots and help you to steadily improve the performance of your system.
Augmentation of rare corner cases
Corner cases are rare but safety-critical scenarios. To show that an automated vehicle is safe, it has to prove its reliable operation in these cases. However, recording corner cases is practically not feasible since they rarely occur in real traffic or they are too risky to stage. This is why we create them by augmenting real data with virtual objects.
Any scenario as realistic as real world data
We exceed the degree of realism of pure simulation. The objects that we insert do not only interact with their surrounding in a physically correct way, they also cast shadows and reflect light from their environment. Our fully automated pipeline allows us to create almost unlimited variations within a scene by varying object types, appearance and behavior.