PROPER HYBRID MODELS FOR SMARTER VEHICLES

The vehicle of the future is “smart”. It is expected that a vehicle can react flexible on changes of its environment and make decisions independently in order to optimally adapt to changing boundary conditions. This requires a high degree of "self-awareness", i.e. the ability to predict the effects of its own behavior in interaction with the environment. Thus, creating such models of vehicle and environment quickly, cost-effectively and in consideration of fidelity and performance is a key competence. Classic model-based approaches are often associated with high development costs. Advances in artificial intelligence open up new options, but are data-intensive and involve other risks. In this project, hybrid (data and physics-based) approaches are to be evaluated in concrete applications in order to be able to generate scalable “proper models” in a data-efficient manner using existing physical knowledge. In future, this will make it possible to develop innovative product properties in a significantly shorter time and to implement them in vehicles.