Scenario Generation

Test and validate your autonomous driving functions with simulatable scenarios taken from the real world.

The test space for autonomous driving functions is incredibly large today and it’s not feasible to validate everything with real-world driving tests.

We bring the complexity of reality into your simulation via highly accurate, complex simulation scenarios and their variations derived from real-world sensor measurements.

Scenario Generation Phases

Scenario Identification

Scenario Identification

The process starts with picking interesting parts out of the petabytes of recorded data collected during real-world test drives. These interesting scenarios can be overtaking maneuvers, unprotected left-turns or pedestrians crossing in front of the car suddenly. These are all challenging scenes for automated driving functions including perception or planning to handle.

Scenario Extraction

Scenario Extraction

With our annotation tools we extract the relevant meta-data from raw data in the next step. Extracting and localizing all the different objects (vehicles, pedestrians, etc.) including their class and trajectory is essential for the creation of a precise digital twin. This so-called “replay” scenario reflects the recorded scenes in a very accurate manner and can be used for simulations. Over time, a scenario database of the right scenarios at the right quality is built.

Scenario Fuzzing

Scenario Fuzzing

In order to generate variations we abstract these scenarios to “logical scenarios”. Which means the trajectories are now represented as distinct maneuvers performed by respective traffic participants like cars, cyclists or pedestrians. Maneuvers are parameterized to allow large scale scenario-based testing on-premise or in the cloud. Parameters are further added to environmental conditions such as the weather and other factors that influence the driving algorithm. They can be varied to challenge the function in different ways. That’s why it’s important to identify relevant, realistic and meaningful parameter ranges.

Challenge your driving functions with edge-case scenarios derived from real-world recordings directly in your simulation. Find and eliminate bugs early in the development process and save precious time and resources. We help you achieve a more complete coverage of the enormous test space through meaningful variations of base scenarios.

Get in touch and request a free demo

Find out more about powerful solutions for the development and testing of autonomous vehicles from dSPACE, the world leader in simulation and validation.

Discover dSPACE