Helmholtz AI project call showcase: AI for the prediction of mobility patterns

Machine learning approaches for large-scale evolving phenomena, such as magnetic field mapping for self-localization, can greatly improve the performance of future navigation tools.

How can machine learning approaches help with vehicle navigation when satellite data is no longer a reliable source? Learn more in this week’s Helmholtz AI project showcase, in which researchers from the Karlsruhe Institute of Technology (KIT) and the German Aerospace Center (DLR) are working on mapping the magnetic field for future mobility and navigation. 

Could you introduce yourself, giving your affiliation, area of work, and of course, the project title?

The 2019 funded Helmholtz AI project ‘Ubiquitous Spatio-Temporal Learning for Future Mobility (ULearn4Mobility)’ is a collaboration between the Intelligent Sensor-Actuator-Systems laboratory (ISAS) at the Karlsruhe Institute of Technology (KIT) (Kailai Li, Florian Pfaff, and Uwe Hanebeck) and the Institute of Communications and Navigation at the German Aerospace Center (DLR) in Oberpfaffenhofen (Benjamin Siebler, Stephan Sand, and Susanna Kaiser).

In simple words, what specifically is your project about? And, how and why do you think it is a high risk, high gain endeavour?

The world we live in is constantly changing. Modelling time-variant phenomena, such as the spread of disease, is crucial in many fields. In ULearn4Mobility, we develop machine learning approaches for large-scale evolving phenomena. Our key application is to learn a map of the magnetic field and keep it updated to utilise it for self-localization for future mobility and navigation within indoor environments. The project is high gain because the ability to localise oneself without satellite signals (which are used in phones and planes) is an important skill. During the Ukrainian war, satellite data was unreliable multiple times, which shows that experts warning to avoid overly relying on this single source of information is dangerous. While we are convinced we will obtain good results, our endeavour is high risk because one never knows if this approach will be one that will come out as the best one in the long run.

How important has the Helmholtz AI funding and platform been to carry out this project?

ISAS and DLR Oberpfaffenhofen have had good contact and joint research interests for years. However, good technical collaborations can only blossom with sufficient funding. Many calls either only make sense at a University like KIT and others mainly for DLR. The Helmholtz AI funding, which works great for both parties, was hence an invaluable tool to start a great collaboration of two parties, which promises to provide better results than any of the two parties could have achieved by itself.

Figure: Shown are intensities of the magnetic field in a laboratory at the DLR.