Research interests

Data-driven validation of AI-based perceptual functions

Automated driving is currently a major vision of the future in the automotive sector. The rapid development in this area is related to the rapid development of machine learning methods. Classical perception functions are gradually being replaced by machine learning and AI functions in the car of the future, as the AI functions are more performant than the classical methods. However, an approved AI function does not yet exist because the safety proof for black-box systems (AI systems) do not yet exist. For this reason, research is being conducted at FZI/ITIV on various methods to test AI functions in a data-driven manner and to evaluate their goodness.

Future software development processes

Established software development models are designed to develop classic software. Classic software is characterized by the fact that everything that is implemented can be tested with deterministic, rule-based and explicit software tests. With AI and machine learning methods entering the vehicle, classical software development models need to evolve. This is because machine learning models cannot be validated by classical tests. At ITIV&FZI, we are researching new software development methods designed for developing and testing AI models trained by data.

Time series analysis

Historical data is increasingly being used in companies to support decision-making. Due to digitalization and the associated easy collection of data, the trend will increase in the future. For predictive maintenance, we are looking at different methods at ITIV/FZI to classify time series and identify patterns in them. This classification can be used for decision making and helps the engineer to make informed decisions.

Offene Studentische Arbeiten
Titel	Forschungsthema	Betreuung
Pattern recognition in time series	Zeitreihenanalyse, Datenanalyse, Mustererkennung	M. Sc. Hanno Stage