Newsletter September 2021 ()

Drone recordings as a basis for simulation and validation of autonomous driving functions

 

IPG Automotive and Automatum enable the utilization of traffic recordings from drones to automatically generate scenarios in virtual test driving

Understanding the behavior of road users is a key factor to validate assisted and automated driving functions. With the help of scenario-based tests, driving functions can be tested reproducibly in relevant and critical situations in the context of the full vehicle. One of the main tasks is to understand, classify and transfer the compiled data to suitable scenarios for virtual test driving.

Ideally, the compiled situations are based on real-world trajectories of a variety of vehicles that drove in the same situation. Drones that record complex traffic situations via top view over a longer period of time, detect road markings as well as objects and define movements in the world coordinate system deliver extensive material. On the one hand, the drone recordings provide a basis for statistically relevant observations of real traffic behavior that are utilized to train agent models for example. On the other hand, the included ground truth data can be used for virtual test driving scenarios. It is therefore possible to link drone recordings and driving simulation even closer on the basis of real-world data.

In this cooperation between IPG Automotive and Automatum, the respective expertise of both companies complement each other: Automatum automates the trajectory generation and IPG Automotive utilizes it on the simulation and test platform CarMaker. The recording of traffic scenarios, detection and classification of the particular objects as well as the description of trajectories are carried out by Automatum. The expert for top view automation also creates the road description with the Scenario Editor from IPG Automotive. According data sets include the road model in the CarMaker data format ROAD5 as well as in an OpenDRIVE file. IPG Automotive uses the trajectories of the classified objects comprised in the data set to create logical, variable scenarios with the toolbox ScenarioRRR (Record-Replay-Rearrange). This allows users to take the ego perspective of every traffic object and to generate according sensor data (lidar, radar, camera, ultrasonic). Hence, top view data sets enable the generation of ground truth data and modeling of the vehicle’s ego perspective.

“For us, it is important to see the direct application of our data.” Paul Spannaus from Automatum derives different benefits from the cooperation with IPG Automotive. “The data sets offer a variety of possibilities for the development process of autonomous driving functions. We are pleased that drone data can be easily integrated in CarMaker and that we can advance the use cases of our data set jointly with IPG Automotive as well as follow their immediate usage.”

Watch the video