The »VMC® Road and Scene Generator« is a software package is currently developed at our institute and enables the virtual development and evaluation of automation systems based on real environment data.
The process works as follows:
- First we select a representative city with traditional statistical methods with regard to selected criteria such as gradient characteristics.
- We then capture it using the institute’s measuring car REDAR and save it as a 3D point cloud.
- We apply methods of Machine Learning and deep learning neuronal networks to analyze and classify the data.
- Relevant objects such as vehicles, lanes, lane markings, buildings, etc. are automatically identified.
- This information provide a key component for a highly accurate sensor simulation since additional attributes can be made available for each object and each point, e.g., material properties, reflection and absorption characteristics depending on various electromagnetic wavelengths, etc.
The data analysis and classification is highly automated leading to a highly efficient overall process.
Environmental Data in OpenDrive® Format and Extended Information
Apart from this procedure, the Road and Scene Generator offers the option of generating and exporting environmental data in OpenDrive® format for individual routes in a specific region exclusively on the basis of VMC®'s own geo-referenced database. This can be done without any additional measurement.
In addition, VMC® then offers the option of enriching exported routes with additional information, such as roughness data, traffic information or driver-, vehicle- and traffic-dependent speed profiles. These data are then used as inputs for test benches or other HiL scenarios and tests.