Motivation and General Goal

Earth observation (EO) has become an operational source of big data. Fostered by the European Copernicus programme with its high-performance satellite fleet and open access policy, the user community has increased and widened considerably during the last years. For example, a recent study by PWC shows that the return of investment into Copernicus is about four euros to one, primarily from its downstream applications (ESA/PWC, 2019). This raises high expectations for valuable thematic products and intelligent knowledge retrieval. In the private sector, NewSpace companies launch(ed) hundreds of small satellites which have become a complementary source of EO data. In the last years IT giants like Google and Amazon entered the scene and brought their expertise of artificial intelligence (AI) to EO data – but often lacked profound EO domain knowledge. The general research goal of the host institutions is to exploit this new and exciting revolution in data-intensive – or even data-driven – science for EO by developing and tailoring novel data science and AI concepts for geo-relevant use cases and combining them with existing rich physical model-based EO expertise. In this context, the specific challenges of EO data must be taken into account, such as heterogeneous data sources, extreme scales, data complexity, shortage of training data and high quality requirements. Also the European Space Agency (ESA) responded to the current disruptive development in EO by founding the Φ-Lab, which intends to accelerate the adoption of AI techniques by EO and space researchers. This vibrant field of AI for EO (AI4EO) will be the home of the proposed innovation lab.

Germany is very well positioned in this blooming field of AI4EO. The number of publications on deep learning (a very recent approach adopted in AI) in EO by German institutions in the past five years ranks number three after China and the US. Among all institutions, the host institutes Technical University of Munich (TUM) and German Aerospace Center (DLR) rank the first among all non-Chinese institutions. Despite these successes, most of the efforts in the international community remain at the exploitation phase through purely application-oriented research. Fundamental science questions remain open. The proposed Future Lab AI4EO (hereinafter referred to as “The Lab”) will bring 20 renowned international organizations in 9 countries and 27 highly-ranked scientists together to research three fundamental, yet rarely addressed, challenges faced by EO-specific cutting-edge AI research, namely, Reasoning, Uncertainties, and Ethics. In addition, other innovative AI4EO topics will be addressed by our Beyond Fellow program, based on which ca. 70 highly talented PhD and PostDoc researchers will invited to visit our lab.

The research carried out in the Lab will not only advance cutting-edge EO science but also lay the foundations, in particular by addressing theoretical analysis, quantification of uncertainties, and ethics of AI, towards equitable AI4EO technology transfer. It will consolidate the pole position of Germany in AI4EO and establish Germany as the AI gravity center in the field of EO.

Social Relevance

While the volume, complexity, and real-time requirements of scientific data across all areas of research are strongly increasing, AI and ML technologies are now in the focus of data science. In the Earth Observation domain, the powerful Copernicus satellites, the huge archive and the free and open data policy offer tremendous opportunities for tackling social grant challenges. For example, to support the ambitious UN SDGs, like creating sustainable cities or to vanish hunger, geo-information retrieval from the steadily growing EO archives raises hope in closing gaps in knowledge and strategic directions. Another example is that it is a global goal to increase biomass and yield production in a sustainable way. To reach this, two measures are most important: irrigation and fertilization. Continuous monitoring of the water availability as well as the crop development and nutrient availability is needed to react quickly and find the most sustainable solutions for the future. For this, intelligent information extraction from EO data is the most promising solution, as it can potentially provide such information that is globally available.

Yet to meet requirements of diverse applications, classic data analytic methods do not suffice anymore. This requires not only new technological approaches to manage and process large amounts of data, but also new analysis methods. This is where AI4EO comes to pay. We need AI4EO. More importantly, High EO quality requirements and wide application diversity call for EO-specific AI research and innovative AI4EO methods.

Thus, cutting-edge EO specific AI research envisioned in the Lab is of great social relevance. Through these novel AI4EO methods, geoinformation derived from big Earth observation data will be invaluable for many scientific, governmental and planning tasks. Geoscience, environmental sciences, cartography, resource management, civil security, disaster relief, smart cities, intelligent transportation, as well as planning and decision support are just a few examples.

Location of the Future Lab AI4EO

The Munich metropolitan region is one of the top places worldwide for AI education and research. With two local excellence universities TUM and Ludwig-Maximilian University (LMU) with their top-ranking computer science and mathematics departments, several research institutions from Helmholtz to Max Planck, and many more, Munich provides an ideal science landscape with already existing collaborations and expertise.

TUM serves as the host institution for the Future Lab AI4EO, represented by Signal Processing in Earth Observation. Consistently ranked among Europe’s leading universities, TUM has recognized the challenges emerging in the digital age and is already spearheading the advancement of Al research from fundamental stages via applied research to the study of Al’s social implications. In this context, the DLR is a strong partner. Together with LMU, Helmholtz Center Munich and Max Planck Institute for Plasma Physics, DLR and TUM established the Munich School of Data Science (MUDS) to train the next generation of “data scientists” – interdisciplinary experts in applying, adapting, and developing methods for AI and data science tailored for a broad array of research domains, including EO. In addition, DLR founded the local Helmholtz Artificial Intelligence Cooperation Unit (HAICU) – MASTr (Munich Unit @ Aeronautics, Space and Transport), which is providing AI expertise from EO, robotics, computer vision and HPC/HPDA support. DLR also started strategic cooperation with Leibniz Supercomputing Centre, e.g. through recently signed cooperation agreement “Terra_Byte”, which shall enable the highly efficient and independent analysis of large amounts of data using the latest methods to understand global trends and their consequences. This further strengthens the already existing network of collaboration.

The Lab will be physically located at the new campus of TUM in Taufkirchen/Ottobrunn, where TUM is currently establishing its new Department of Aerospace and Geodesy as part of the Bavarian space initiative. Besides TUM, the area also hosts the University of Federal Armed Forces, Munich Aerospace, and several EO and space industry partners (e.g. Airbus, Siemens, and IABG), thus providing the ideal location for knowledge and technology transfer. In addition, two further institutions, namely Munich Data Science Institute (MDSI) and the TUM Institute on Ethics in AI, are strong partners, who will provide data science theoretical support and ethical guidance to AI4EO.

Link:

– TUM-SiPEO: https://www.sipeo.lrg.tum.de
– MUDS https://www.mu-ds.de/
– Helmholtz AI – Local Unit at DLR: https://www.helmholtz.ai
– Department Aerospace and Geodesy: https://www.lrg.tum.de
– TUM Institute for Ethics in Artificial Intelligence: https://ieai.mcts.tum.de
– TUM Data Science Institute: https://www.mdsi.tum.de/mdsi/startseite

Funding Program

Förderung von internationalen Zukunftslaboren in Deutschland zur Künstlichen Intelligenz

Artificial Intelligence for Earth Observation