Overview: Today, hundreds of small applications run in modern cars that ensure roadworthiness with the help of sensors and control logic. These applications have special requirements with regard to real-time capability and reliability. To ensure these two basic conditions for the software, complex mathematical calculations are performed to determine the feasibility of the tasks. In addition, the final product has to prove the scarcity of occurring errors in field tests. Since the trend goes towards a consolidation of the software modules on only a few, but powerful, devices, a failure of these or a mutual blocking of the software modules has far-reaching consequences. In order to counter these increasing problems, the safe and efficient execution of the individual tasks must be guaranteed. Due to the increasing size and complexity of the code base in vehicles, this can no longer be achieved by mathematical and platform-specific approaches alone. Therefore, it is investigated whether AI-supported approaches, such as machine learning, are suitable. Machine learning, in turn, requires a solid and reliable data set as a basis, which is the subject of this thesis.