Maschinelles Lernen für die echtzeitfähige und hochgenaue Lokalisierung mit Multisensorsystemen: Inertial Measurement Unit (IMUs) like accelerometer and gyroscopes are becoming cheaper and cheaper, as they are increasingly used in consumer applications. However, for these applications, short time stability is mostly sufficient, so that these devices show a large dependency (drift) on time and temperature.
Based on previous analysis, it can be assumed that ML-based support of calibration processes might significantly improve the resulting accuracy of these low cost devices also for long-term industrial applications. It is the student’s task to analyze the results until now and to propose ML-algorithms for a “self-learning” calibration. Based on the results, experiments shall be conducted and the proposed algorithms shall be verified.