Autonomous Driving

Correctly assessing risks and predicting the driving situation are key to safe and comfortable driving. This applies to conventional vehicles, which are equipped with e.g. distance and lane keeping assistance systems and traffic jam pilot, as well as to fully autonomous driving vehicles.
‍
The AI trained and implemented by EDI as a solution for autonomous driving uses Dynamic Risk Management (EDI-DRM) developed by us. In order to develop an appropriate speed and strategies for different driving situations, the EDI-DRM includes the current volume of traffic, the weather conditions and vehicle-specific states in the calculations and also adapts them to the existing infrastructure.
Depending on the external conditions, the state of the vehicle and the wishes of the passengers, the appropriate speed and the correct safety distance are essential as a "safety cushion" to reach the destination safely. By combining DRM and safety cushion, EDI provides different mobility solutions for autonomous driving and safe navigation.
‍
As part of the AnRox research project, we are involved in the development of an autonomous driving robot taxi together with Bosch, Mercedes Benz, Siemens, Infineon and many other partners. Our contribution is a smart intelligence layer with the EDI-DRM. In the event of critical system states of the vehicle, this intelligence layer designs a solution strategy depending on the driving situation so that the passengers of the robot taxi are safe at all times. Adapted to the respective traffic situation, different strategies are used depending on the route.
‍
Efficient development and introduction of automated driving functions and algorithms through virtual certification (TÜV)
‍Our incident detector application can automatically evaluate critical scenarios in traffic data and weak sensor signals of the vehicle, e.g. of the cameras, and transfer them into a simulation file (OpenScenario). This enables the development and testing of automated driving functions and algorithms based on real scenarios in road traffic in multiple real time and in great variation. For this purpose, both the driving and sensor behaviour, as well as the behaviour of other road users, are evaluated using a combination of different algorithms from the field of machine learning (autoML). The results are virtual training & testing scenarios for testing the driving behaviour of automated driving functions and autonomous driving vehicles.