SaiNSOR
SaiNSOR is an enabler project that develops fundamental technologies and methods and translates them into specific measurement systems for cross-programme applications. The focus is on the close integration of sensor technology and data processing using artificial intelligence (AI) methods.
The development of new, application-specific data-based methods for extracting information (e.g. using machine learning approaches) depends largely on the reliable provision of high-quality data. The potential of modern methods for analysing and interpreting large amounts of data with a high information content (AI methods in the broadest sense) is well known. Operational real and virtual sensors, standardised interfaces and software and hardware modules for data pre-processing are required to generate DLR-specific data for AI methods. These elements are the backbone for the development of intelligent sensor technology in general and sensor-based AI in particular (e.g. linking physical and data-based models, edge computing)
The Institute of Data Science is specifically represented here in the work package on edge computing. In practice, there are often few computing resources available for recording and processing data close to the sensor. This places special demands on AI methods. In detail, the efficiency of machine learning methods for anomaly detection or uncertainty estimation on various edge devices will be investigated and analysed as part of SaiNSOR. The findings from these experiments will then be used to adapt and further develop these machine learning methods for effective use on edge devices.
Furthermore, AI methods in edge applications are subject to increased attacks due to their special exposure. Using real-time monitoring methods and addition tools from software security, we ensure smooth operation and detect attacks before they can affect the rest of the system.