29 Sep Testing Automated Driving Beyond Perfect Conditions
Autumn is approaching – and with it come wetter roads, rain, spray and increasingly unpredictable weather. Conditions like these are challenging enough for human drivers. But how do automated driving systems perform when sensors can no longer rely on perfect visibility?
This question was at the heart of ADApproved – Autonomous Driving Sensor Homologation Lab, which has now been successfully completed after three and a half years of research and development.
Creating Realistic Test Conditions
Within the project, the existing test infrastructure at the Mobility & Sensor Center in Roding was significantly expanded. Today, a much broader range of driving scenarios can be reproduced under controlled and repeatable environmental conditions. Tests can be carried out both physically in the indoor test facility and virtually using a digital twin, allowing real and simulated testing approaches to complement each other.
Among the key developments are an enhanced spray demonstrator, improved rain-related machine learning models and the integration of indoor localization. The project also enabled repeatable Automatic Emergency Braking (AEB) tests, including the collection and evaluation of reference data. This makes it possible to systematically assess and compare different vehicle systems under standardized conditions.
SETLabs’ Role in ADApproved
As consortium lead, SETLabs coordinated ADApproved and brought together the expertise of the project partners, including Technische Hochschule Ingolstadt and AVL Software and Functions GmbH.
SETLabs developed a data-driven machine learning method to simulate the effects of rain on LiDAR intensity data. The voxel-based method uses a CNN U-Net architecture and was demonstrated within the CARLA simulation platform. The configurable module can also be integrated into other simulation tools and frameworks.
From Research to Real-World Application
The practical potential of the infrastructure is already evident. AEB benchmark activities have generated strong interest from industry, and the test environment has also been used in measurement campaigns involving ADAC and the German Federal Highway Research Institute (BASt).
Towards More Robust Automated Driving
For SETLabs, ADApproved demonstrates how physical testing, virtual methods and intelligent software can be combined to make automated driving systems more robust, comparable and ready for real-world conditions – including the ones that are anything but perfect.
ADApproved
Autonomous Driving – Sensor Homologation Lab
Duration: 01/2023 – 12/2025
Project volume: EUR 300.000
Funding: 100 %
Funded by the „Bayerische Staatsministerium für Wirtschaft, Landesentwicklung und Energie“ within the framework of the Bavarian collaborative research program (BayVFP) in the funding line „Digitalisierung”, Förderbereich „Informations- und Kommunikationstechnik (IuK)”.
Supported by Bayern Innovativ – Bayerische Gesellschaft für Innotvion und Wissenstransfer mbH