Virtual Validation through Varied Reality
The EU-funded project Vivare unites the Intenta GmbH and the the chair Process Automation of the Chemnitz University of Technology in solving the problem of the constantly growing demand for data in the validation of driving assistance systems.
The central problem is the infrequent occurrence of situations that are interesting or even critical for validation.
In general, 30.000 to 60.000 miles elapse between two activations of the first tier of an emergency brake assistant during a set of test drives.
The second stage, which is much more important and much more critical than the first, did not activate once during that time.
The goal of this project is to improve the validation of driving assistance systems by enhancing real test drives with simulated road users.Simulated road users help to increase the amount of critical traffic situations and thus reduce the amount of required data.
Neuronal networks are capable to detect road users by using data of many different sensors.
Hence they are ideal to simulate test drives. The data is used to generate traffic scenarios which can be customized using an editor. In the edited test drive, realistic high-level sensor outputs are generated for the "virtual" vehicles.
Data fusion from point cloud and camera image
Data Processing and Software Development:
- automatic evaluation of sensor data
- tools for sensor data annotation
- hypotheses generation for supporting data annotation
- development of generative sensor models
- editor for variation of traffic scenarios
- generation of sensor data for virtual road users
Our partners at the Chemnitz University of Technology supports us in the development of neuronal networks.
- significant reduction of required data for validating driving assistance systems
- thus enabling the validation of more complex systems
- safe validation of critical traffic situations