Name of the participant: Oliver Bleisinger
Description of the IT research project: The increasing networking of various IT systems and the digitalization of test benches and vehicle tests are increasing the availability of measurement data in vehicle development. For example, artificial neural networks can be used to perform physical simulations to support various processes in virtual engineering. In addition to the use of digital twins and machine learning in the engineering of vehicle systems, it is also possible to train predictors at low cost, which can be used for the control of complex processes or for the department of recommended actions in autonomous driving.
The project will address the following questions:
- Which correlations of vehicle dynamics can be learned in principle and which learning methods are suitable for AI-based simulation in this context?
- Which data basis is required and which model accuracy can be achieved for AI-based simulation in the field of vehicle dynamics/engineering?
- What do suitable software modules or algorithms look like to support the learning of simulation models and how can they be prototyped?
- Can the researched approach be integrated into concepts of model predictive control in highly automated and autonomous driving and increase the control accuracy?
The aim of the AMOSCO project is therefore to find out to what extent and on what data basis AI-based simulation is suitable in the field of vehicle dynamics. The vision: In the future, simulation models will no longer be created manually, but will be learnable at the push of a button using vehicle measurement data and innovative IT tools.
Software Campus partners: Fraunhofer IESE, IAV
Implementation period: 01.08.2021 – 31.07.2023