Validating Sensor Models

• Using generated scenarios to validate sensor models
• Realistic sensor simulation
• Test scenarios on the basis of real measurement data

Task: Sensor Model Validation

Validating sensor models by using generated scenarios

Challenge: Ensuring Accuracy in Sensor Simulation

To be able to validate the perception algorithms by means of simulation, the sensors must be simulated with a high degree of precision and cover all relevant effects of the real scenarios.

Solution: Test Scenarios on the Basis of Real Measurement Data

To achieve the required level of precision, the sensor simulation is compared with real measurement data. For this, the simulated and the real sensors must be used for the same scenario. With the scenario generation service from dSPACE and understand.ai, measurement data can be used to create relevant simulation scenarios. The scenarios can accurately simulate the static and the dynamic environment, thereby providing a suitable environment model for the physical sensor simulation.

Benefits of Scenario Generation by dSPACE and understand.ai

In a first step, a precise 3-D model of the environment is created that enables physical sensor simulation. The dynamic parts of the scenario (moving vehicles, pedestrians, etc.) are reproduced with high precision so that the objects are at the correct position at any given time.

Transferring the sensor principles: The structure of the created 3-D models make them usable for all relevant sensor types. For example, a scenario that was created using lidar data can also be used for camera or radar simulation.