German Research Center for Artificial Intelligencenews 01 I 2022 Download this issue 9 4 e u s s i I K F D 2 2 0 2 © / 1 5 2 2 - 6 9 1 2 - n s s i Industrial AI Key to Resilience and Flexibility in Globalized Value Chains German Chancellor Olaf Scholz Visits DFKI AI for the Intensive Care Unit Kaiserslautern Site Real-time Vital Data Classifica- tion Relieves Medical Staff and Improves Treatment Quality

INDUSTRIAL AI, TRUSTED AI, AND 03 L A I R O T I D E 2 2 0 2 / 1 s w e N I K F D More than two years of the Corona pandemic, the blockage of the Suez Canal by the container ship Ever Given, the war in Ukraine, and the ensuing energy price crisis have shown how vulnerable supply chains can be in our globalized industrial world. Major supply interruptions can bring individual sectors and entire branches of production to a standstill, jeopardize order processing, and lead to drastic losses in sales. Industrial AI (p. 6) is the key to more flexible manufacturing processes and dynamic adaptation to ever-changing production conditions. AI technologies are crucial for three of the core concepts of Industrie 4.0 – digital twins, decentralized control, and active product memory. The best way to strengthen the competitiveness of the manufacturing sector and to maintain quality leadership is with industrial AI. AI creates new added value: for example, incidentally collected data on raw materials in the food industry serves to enable the semi-automated generation of higher-value data products through decentralized AI-Services (p. 10). In the field of agricultural food production, Nature Robots, a DFKI spin-off, is developing software components that enable robots to independently navigate between crops and weeds (p. 22). AI, which was once a niche science, has been transformed into a “mega-technology” to benefit the whole of society through machine learning with artificial neural networks. However, the traceabili- ty and reliability of the results, as well as the systems‘ lack of a self-explanatory capability, are prob- lematic. The general need is for AI systems that can be trusted. Trusted AI (p. 14) is the term that refers to a new generation of AI systems that provide guarantees about their functionality, allowing use in critical applications and paving the way for the development and use of long-term autono- mous systems (p. 20). AI is assisting nurses in the intensive care unit. The service consists of holistic real-time analysis and classification of the patient‘s vital signs, which are continuously monitored. AI-powered systems detect critical conditions at an early stage, generate alerts, and serve to actively improve patient safety. The European AI Centers of Excellence network, which is designed to be the nucleus for further activities in cutting-edge AI research and innovation in Europe, is being significantly shaped by DFKI in the VISION, TAILOR and HumanE-AI-Net projects (p. 16). The collaboration between DFKI and Inria is also being steadily intensified: R4Agri and ENGAGE are two new projects started at the beginning of this year as part of the German-French AI cooperation (p. 33). Prof. Dr. Antonio Krüger Helmut Ditzer Prof. Dr. Prof. h.c. Andreas Dengel Prof. Dr. Dr. h.c. Frank Kirchner Prof. Dr. Philipp Slusallek

2222 e v a w r e v e © 882828 30 Minister President Malu Dreyer Visits DFKI Branch Office in Trier 30 Project AScore – Simulation-based Pandemic Management System for Municipalities Successfully Closed-out Imprint Issue 49, May 2022, ISSN 2196-2251 Published by German Research Center for Artificial Intelligence GmbH (DFKI) Editorial staff Heike Leonhard, Christof Burgard, Reinhard Karger, Armindo Ribeiro, Stephan Lehberger Editorial address Saarland Informatics Campus D3 2, Stuhlsatzenhausweg 3, 66123 Saarbrücken, Germany E-Mail news@dfki.de Phone +49 681 85775 5390 Translation team Glenn Peach, Armindo Ribeiro, Sylvia Krüger, Heike Leonhard Photo credits DFKI, unless otherwise noted Layout, Graphics Christof Burgard Production One Vision Design Responsible Heike Leonhard, M.A. Frequency of publication Semi-annual News online www.dfki.de/en/dfki-news 3323232 343434 05 T N E T N O C 2 2 0 2 / 1 s w e N I K F D t n e t n o CC C

07 2 2 0 2 / 1 s w e N I K F D 76.2 percent of the machine manufacturing and plant engineering companies identified with the statement, “It is part of their professional philosophy as technolo- gy leaders to address this topic,” according to a study at the time by the Impuls Foundation. Three core concepts of Industrie 4.0 are critical – the dig- ital twin, distributed control, and active product memory. In 2011, the authors wrote, “These concepts make it possible, for example, to better meet the economic as well as the special ecological requirements of “green production” for a CO2-neutral, energy-ecient city.” It all started with product memory DFKI was addressing the topic of product memory in the SemProM project already back in 2008. The next gener- ation of mobile, embedded, and radio-based elements for semantic Internet communication between everyday objects was studied as part of the ICT-2020 research pro- gram of the Federal Ministry of Education and Research (BMBF). The goal was to enable a blank disk to access all existing operational and machine data to document the entire manufacturing process. Since the passage of the ‚Corporate Due Diligence in Supply Chains Act’ by the German Bundestag in June 2021, product memories have taken on increased importance and new meaning as they can provide an important building block for com- panies to meet the legal obligation to provide evidence. The ‘Supply Chain Act,’ which becomes effective on January 1, 2023, will initially apply to companies with more than 3,000 employees, but will be extended to compa- nies with 1,000 employees beginning in 2024. Product memory is actively discussed in terms of In- dustrial AI and is turning conventional production on its head. This leads to a rethinking of control logic: tra- ditionally, the rate, the work plan, and a central control would govern. Industrial AI enables self-adapting process control in place of isolated corrections to the machine schedules. In a tightly networked production system, the blank disk directs the manufacture of what will become the finished product. This not only enables distributed production control of individual products and very small batches but also documents resource consumption, i.e., the materials and energy used for each individual work- piece. This implies the need for seamless networking and closely detailed information exchange in the Enterprise Resource Planning (ERP) and Manufacturing Execution Systems (MES), the major systems used for corporate planning and factory control. In multimodal teach-ins, robots learn movement patterns that they can subsequently execute independently.

09 2 2 0 2 / 1 s w e N I K F D AI services in production are designed to avoid downtime and reduce scraps. The importance of Industrial AI In Germany, the industry’s share of gross value added is 22.9 percent, the highest of all G7 countries. The strongest sectors are automotive construction, electrical engineer- ing, equipment manufacturing, and chemicals. Accord- ing to KI-Zukunftskompass, a survey commissioned by Bosch in 2020, a majority of Germans (53 percent) think the use of artificial intelligence is necessary if companies are to remain competitive in future international competition. The German economy is char- acterized by high-quality in- dustrial centers, a strong export industry, and a pow- erful SME sector. Industri- al AI provides the best op- portunities for strengthening its digital competitiveness and maintaining its quality leadership k i p e e r f / r o t c e v o r c a m © into the future. semi-finished and finished products is possible. Actions can then be recommended to optimize the parameters for the production machinery as well as to determine the quality of semi-finished and finished products. By avoid- ing production downtime and product rejects, this con- tributes to reducing failures and increasing cost savings. Services to increase resilience are offered to companies on an open digital platform – without having to build up extensive AI expertise themselves or hand over their data sovereignty. Specifically, these requirements are implemented in the manufacturing and process industry using AI-based anomaly detection and the concept of the digital twin. This digital representation models all the possibilities of an industrial process chain so that all possible effects of a failure can be simulated. The digital twin The digital twin can do even more. The concept applies not just to simulations but also as a real-world input in- terface. In human-robot collaboration, for example, it enables seamless collaboration between workers and modern cobots. In contrast to convention- al industrial robots, which have to act separately from humans for safety reasons, the collaborative light- weight cobots provide direct assistance to humans. Areas of application in manu- facturing include, for example, the assembly or the placing and removal of the workpiece to and from the processing machine. It is also suited for use where flexible scalability is desired in the degree of production automation, for example, to cover peak capacity flexibly. When fed with additional sensor data, the digital twin of a production cell becomes the input interface for a ro- botic teach-in on the factory floor – regardless of where it is located. DFKI and ZeMA, as partners in the RICAIP Research and Innovation Center for Advanced Industri- al Production, are developing a collaborative robot arm that can be intuitively controlled remotely in real time by a person using a virtual reality headset. The tech- nologically complex scenario enables close interaction between distant locations and can be used for remote maintenance or remote manufacturing.

11 2 2 0 2 / 1 s w e N I K F D guarantee ownership and control of data assets as well as the execution of defined taskings and sharing of results ac- cording to contract. Use case: Dry beans Dry beans are the most widely produced product in the category of edible legumes worldwide. As part of an on- going trend toward vegetarian and vegan diets, beans not only help support a low-impact lifestyle but also contrib- ute to increased energy eciency and climate protection through reduced use of nitrogen fertilizers and CO2 savings. The many varieties of dry beans exhibit a wide genetic di- versity and quality, which has a decisive influence on crop production. A dry bean classification system is essential for both production and marketing to ensure the principles of sustainable agriculture are observed. Let’s say Farmer A wants to classify his dry beans to produce and sell uniform seed varieties at optimal prices. The farmer meets Farmer B and concludes an electronic contract via eContract Broker. The object of the contract is the result from a joint analysis of the data sets to determine an im- proved classification. Both farmers transmit their dry bean data sets with 16 characteristics per variety to the broker. The broker then executes the analysis and generates a vari- ety classification using neural networks on Edge. The broker sends the result to Farmer A, and both data sets are delet- ed. Farmer A pays Farmer B for his participation and can use the results to optimize his seeds and determine the optimal price for his crop. The eContract Broker also receives a fi- nancial payment for ensuring sovereign, shared data anal- ysis as well as for executing the contract. PROJECT DESCRIPTION Partners: • German Research Center for Artificial Intelligence, DFKI (lead manager) • Agricultural Market Information Company (AMI) • Lindt & Sprüngli Group • Research Institute for Rationalization (FIR) at RWTH Aachen University • Software AG • Saarland University Associate Partners: BVE e.V. DIL e.V. Funding agency: Federal Ministry of Economic Affairs and Climate Action (BMWK) Sponsor German Aerospace Center (DLR) Volume 2.3 million euros The consortium‘s successful submission of the “FAST” GAIA-X use case, which deals with data interoperability and data sovereignty in the food production industry, is attributable to these technical results. The EVAREST consortium also pub- lished DIN SPEC 91452, “Business Model Development for a Data Ecosystem in Food Production.” The standard provides a design approach for specifying business models in data eco- systems and for developing and building data ecosystems. Contact Hannah Stein Research department Smart Service Engineering hannah.stein@dfki.de +49 681 85775 5270 The EVAREST project team at the closing event on April 29, 2022, at the Digital Technologies Forum in Berlin.

13 13 13 I I K K F F D D 2 2 0 2 / 1 s w e N I K F D 2 2 2 2 0 0 2 2 / / 1 1 1 S S W W W E E N N I I K K F F D D LIDARSHARED: THE ROAD TO Rather than competing for the scarce public space in the city, future motorists will share it in “shared spaces.” Shared spaces pose a challenge for autonomous drive vehicles, mainly because collision avoidance systems employed in dense, unregulated, and highly dynamic areas ask a lot from current AI algorithms. Continued research and development of special AI methods is an essential step towards ensuring a high level of safety for all road users and sustainable future transportation. team had to synchronize all sensor data and convert GPS coordinates into positions on the local, campus-wide map. The methods and datasets developed in the LidarShared project have become baselines for future projects in the field of robust and trustworthy computer vision and auton- omous driving. The researchers have already begun to ex- ploit their experience in follow-on projects, paving the way for safe and reliable AI-based systems with little or no hu- man supervision. A presentation of results to the public is planned for summer 2022. Contact Farzad Nozarian Research Department Agents and Simulated Reality farzad.nozarian@dfki.de +49 681 85775 5462 The LidarShared project team at DFKI‘s Agents and Sim- ulated Reality research area (ASR) uses AI algorithms in combination with solid-state Lidar sensors to bring auton- omous navigation in dense and crowded shared spaces one step closer to practical reality. The idea is to use an AI research framework based on open source operations for various computer vision tasks to reliably detect road users in camera images and Lidar point clouds, and then contin- ue to track them through long enough periods to predict their future routes. A combination of different methods and concepts is employed to realize the project goal – a robust and highly reliable collision avoidance system for autono- mous navigation. The tools include unsupervised learning, knowledge transfer, domain adaptation, and sensor fusion. The research team develops synthetic data capture, labe- ling, and visualization tools to create a multi-sensor, syn- thetic dataset for 3D object detection based on a real-world autonomous driving dataset. To better represent real-life situations, the team improved the quality of the synthetic dataset by carefully fine-tuning the parameters of the sim- ulated sensors and environments. “To collect real-world data, we equipped a test vehicle with a front-facing RGB camera, lidar, IMU, GPS sensors, and an embedded processing unit and captured the data on the Campus of Saarland University,” said project manager Farzad Nozarian. We also developed tools to convert raw data into the pub- lic KITTI format for the Lidar odometric data and the 2D/3D object detection and tracking data. In addition to calibrat- ing the camera and IMU sensors, estimating camera-lidar calibration parameters, and rectifying camera images, the An AI framework is to reliably recognize trac participants in lidar point clouds and camera images.

15 2 2 0 2 / 1 s w e N I K F D forts to strengthen digital competitiveness under a com- mon and visionary goal. “Trusted AI Made in Germany” has the potential to become a globally visible brand that car- ries Germany‘s claim of industrial quality leadership into the digital future. More information https://tailor-network.eu Contact Dr. André Meyer-Vitali Research Department Agents and Simulated Reality andre.meyer-vitali@dfki.de +49 681 85775 7241 m o c . e b o d a . k c o t s - t e r o c e d s © The EU‘s High-Level Expert Group (HLEG) defined princi- ples in the publication “Ethics Guidelines for Trustworthy Artificial Intelligence.” AI applications of the future must be legal, robust, and respectful of European ethical prin- ciples and values. Legislation obliging these requirements will come into force in the next few years. The requirements are particularly strict when it comes to applications with significant physical, economic, or social risk. The AI systems used in such applications are assumed to have been validated and certified. Hybrid AI provides ex- actly the greater transparency that is necessary. Hybrid AI approaches are studied and applied by the Agents and Simulated Reality research area (ASR) at DFKI, where a newly developed system of possible combinations is help- ing to assess the advantages and disadvantages in different areas of application. The main research goal of these ef- forts is Trusted AI. Current research is focused on the area of safety engineering as well as various aspects of valida- tion and certification of AI systems and decision making in human and AI agent teams (Human Empowerment). Trusted AI implies that trust and reliability can be reformu- lated as a value proposition. Every economy and society will have to deal with the challenges and threats described above in the near future. Those who manage to push the boundaries of controllability for AI will gain a fundamen- tal competitive advantage. Especially for Germany and Europe, this provides a great opportunity to pool their ef- m o c . e b o d a . k c o t s , s k r o w r o m a t e m ©

VISION – Coordinating the European AI network 17 2 2 0 2 / 1 s w e N I K F D The objective of the VISION project is to coordinate and align the work of European networks of AI Excellence Cen- ters: AI4Media, ELISE, Huma- neAI Net, and TAILOR, as well as a vibrant exchange with the European Commission. The key role assumed by DFKI is to organize the collaboration between research and industry. www.vision4ai.eu TAILOR – Trusted European AI At present, 55 partners are working at the European level with representatives from research and industry in TAILOR to develop the scientific basis for Trusted AI. Approaches and methods from different areas of AI, such as machine learning, are integrated in meaningful combinations to address the components that are important for trusted AI, such as explainability, security, fairness, liability, responsibility, reproducibility, data protection, and privacy, as well as sustainability. https://tailor-network.eu HumanE-AI-Net – Human-centric AI HumaneAI Net brings together top European research centers, universities, and key industry players within a Centers of Excellence Network. The network combines world-leading AI competence with key players in related fields such as Human-Computer Interaction, cognitive science, and social sciences. www.humane-ai.eu More information https://claire-ai.org Contact Dr. Silke Balzert-Walter Research Department Agents and Simulated Reality Claire Oce Germany oce-germany@claire-ai.org +49 681 85775 2107

19 2 2 0 2 / 1 s w e N I K F D The centerpiece of the program was “Human-Centric AI” and successful transfer to business and society. in cooperation with the federal government, the states, our shareholders, and industry partners, and we explained our ambitious plans to continue shaping this path in the future.” Prof. Dr. Andreas Dengel, Scientific Director and Site Man- ager at DFKI-Kaiserslautern and AI ambassador for Rhine- land-Palatinate, said, “The integration of beneficial tech- nologies into society relies on the sustainable transfer of research results. A successful transfer is not a one-way street but depends on a vital ecosystem of research and applications and close cooperation. Today, using selected practical examples, we showed how artificial intelligence can improve people‘s cognitive work and how AI can help to master major societal challenges such as a pandemic and climate change. The visit by the Federal Chancellor, the Minister President, and the Science Minister under- lines the relevance of AI research. It expresses a great appreciation of the research and transfer achievements of our employees, not just at the Kaiserslautern site, but all of DFKI.” Third Mission: The DFKI transfer ecosystem The presentations included DFKI’s transfer ecosystem, represented by selected technological advances from the Kaiserslautern transfer labs. Special transfer labs provide custom plans for companies of all sizes to enable partners outside of the funded con- sortium projects to continuously and actively participate in current AI developments. The labs follow a proven suc- cess model for the direct transfer of research results to private industry: customer profiles and wishes are incor- porated into the targeted design of innovative solutions from the broad field of AI research. Over at the AI4EO Solution Factory, DFKI AI expertise is combined with Earth observation data from ESA to de- velop new business models and products. Currently, the largest project in this is the Yield Consortium transfer lab. The aim is to develop better yield prediction and risk models using AI to estimate the condition and growth trajectories for all important crops. The SAIL/Sartorius transfer lab is concerned with develop- ment and use of AI tools for biopharmaceutical products. Researchers here study Deep Learning algorithms and image recognition of cells and organoids for the analysis and modeling of biological systems and for simulation and optimization of biopharmaceutical production processes. The transfer lab known as AIforPol focuses on police work with state and federal authorities (BKA/LKA), researching and testing possible AI applications in the area of inter- nal security. Intelligent solutions are required, in particu- lar, for the analysis of the data associated with evidence gathering and administration, as well as for the recogni- tion of specific patterns in mass data. Pandemic simulation for municipalities and smart sensor systems for manual labor Other highlights included two practical demonstrations: The AScore Pandemic Cockpit facilitates local level de- cision-making about Corona measures. The intuitive dashboard evaluates the data used to estimate the im- pact of specific measures at the local level using simu- lation software. AScore: The Pandemic Cockpit for municipalities allows crisis staff to simulate the specific outcomes of local measures. The BIONIC project demonstrates an intelligent sensor network for reducing physical stress. Objective, real-time motion analysis is performed by mobile secure data sys- tems that are robust and user-friendly to alert people to their poor posture. The technology recently made its way from research to market with the spin-off Sci-Track. Contact Christian Heyer Head of Communications Department DFKI Kaiserslautern christian.heyer@dfki.de +49 631 20575 1710

two-armed AUV that operates both autonomously and remotely with two integrated manipulator arms for com- pleting maintenance tasks and repairs to underwater structures. The partners created a powerful IT infrastructure that en- ables intuitive control and monitoring of the underwater robot. Furthermore, this infrastructure ensures a reliable, bi-directional flow of information with the facility operator and allows integration with existing business processes. The AUV in the water column in vertical mode with manipu- lator arms extended. The AUV performs manipu- lation tasks in the vertical mode on an underwater mockup. In Mare-IT, the DFKI Robotics Innovation Center success- WITTENSTEIN Cyber Motor is the company that devel- fully designed and built the “Cuttlefish” AUV, an inter- ops the highly robust thrusters that power the AUV‘s 21 2 2 0 2 / 1 s w e N I K F D vention robot that can be freely positioned in the water column. On the ventral side of this innovative, deep sea going vehicle are two gripping systems that enable it to manipulate objects under water. It is capable of changing its center of gravity and buoyancy during a dive and can assume and maintain any orientation in a stable manner. Besides a fully autonomous mode, the robot can be oper- ated in a hybrid mode using an optical fiber which allows human intervention, for example during critical opera- tions on underwater structures. m o c e b o d a k c o t s - t e r o c e d s © movement. These enable the high-precision maneu- vers required for docking with underwater structures. . SAP contributes to Mare-IT by investigating how to best integrate autonomous systems like AUVs with the . IT infrastructure. To ensure interoperability and reduce integration time and costs, the company implemented a manufacturer-independent integration of a digital twin into the business applications by specifying the administrative shell. ROSEN Group developed four dierent measuring technologies in the project for AUVs to facilitate checking the condition of various underwater structures: sensors detect ferromagnetic structures and measure the electric fields of cathodic protection systems, while other sensors aided by the manipulators on the AUV measure ultrason- ic and eddy currents to determine local wall thicknesses. The project Mare-IT was funded by the Federal Ministry of Education and Research under the grant number O1lS17029A. A researcher monitors the AUV‘s collision avoidance in dual-arm mode from the edge of the pool DFKI Cognitive Assistants research unit has developed a virtual co-pilot that supports the human operator at the control station with teleoperation tasks. The operator is always informed about the current status of the AUV via a holographic, 3D representation on the display thanks to a Microsoft HoloLens. The measured values and sensor in- formation are displayed and warnings are configured via natural language interaction. The AUV and its cameras More information https://robotik.dfki-bremen.de/de/forschung/ can also be controlled via voice commands and, in com- projekte/mare-it bination with eye tracking, information about specific el- www.youtube.com/watch?v=-NcroB_OaW4 ements of the infrastructure is retrievable. Supplemental to the innovative interactive systems, a web application Contact developed at DFKI‘s Embedded Intelligence unit allows robot control and parameter settings via mobile devices such as smartphones and tablets. Dr. Leif Christensen Research Department Robotics Innovation Center leif.christensen@dfki.de +49 421 178 45 4149

s g ü ü t n ö ö r G © 23 2 2 0 2 / 1 s w e N I K F D A 3D point cloud of lettuce in color captured by a robot while driving. The point clouds are used to analyze the crops. A Grööntüügs market garden in Ibbenbüren, North Rhine- Westphalia, from above. The promising principle of market gardening is to be advanced by AI applications. Theoretically, any city or town can grow its own fresh veg- etables through market gardening.” Plagemann agreed to make his garden plots available to the team for test drives with the robots and to contribute his expertise to the de- sign of an expert system. “Market gardeners look for ways to decrease the likelihood of something going wrong. As your own experience grows, you create more favorable conditions in the growing plots. Working with high-tech development that supports this is an important step to- wards smart farming sector-wide.” RESEARCHER AND FOUNDER In December 2021, Sebastian Pütz successfully defend- ed his dissertation entitled “Navigation Control & Path Planning for Autonomous Mobile Robots.” The examining committee members were Prof. Dr. Joachim Hertzberg (Osnabrück University, DFKI) and Prof. Dr. Nick Hawes (Oxford Robotics Institute). Pütz studied Computer Sci- ence at Osnabrück University and did the research for his Ph.D. at the University of Auckland, New Zealand, and the University of Oxford, UK. He is employed as a researcher at DFKI in Osnabrück and manages the PlantMap spin- off, Nature Robots. Pütz can also imagine the technology of his research area being applied to viticulture or agro-forestry. Time will tell if his vision becomes a reality in the future. At present, most startups in countries like the USA and France focus main- ly on conventional agriculture. Prof. Joachim Hertzberg, DFKI Laboratory Niedersachsen spokesperson, supported the cause from the start: “No business can be created from a distance. AI solutions must spring from the application area where they are needed. Here, in the Osnabrück region, everyone involved in agri-tech is within reach – agricultur- al machinery manufacturers, agricultural scientists at uni- versities, garden centers, and farmers. I am excited to see how Team PlantMap will take advantage of this to build the right applications on a strong technological foundation.” More information https://naturerobots.de Contact Dr. Sebastian Pütz Research Department Plan-Based Robot Control sebastian.puetz@dfki.de +49 541 3860502262 The project team of DFKI (from the left): Lars Schilling, Gerrit Woeckner, Sebastian Pütz, Sven Lake, Malte Kleine Piening.

25 2 2 0 2 / 1 s w e N I K F D e v a w r e v e © The Moringa is equipped with a camera mounted above the conveyor belt. As in the earlier airborne projects, this generates overhead image data of the waste collected and sends it live to a laptop in the ship‘s cabin. There, the data is subjected to real-time assessment using AI tools. “Here again, we use machine learning methods. We train the system to differentiate the various quantities of litter on the conveyor belt by type,” said DFKI researcher Wolf. The report does not identify just plastic or organic wastes, “but rather an exact count of how many thousands of bottles, cups, bags, metal cans, or whatever else was brought on board,” said Clemens Feigl from everwave. The findings have a double benefit: First off, the on-site waste management company uses detailed figures about the cargo to determine whether to dispose of it or, in the best case, place it back into the life cycle. Secondly, it facilitates a targeted approach to education and envi- ronmental training in places where a particularly large amount of plastic is produced, for example, at a public festival when hundreds of thousands of disposable cups are used and not disposed of properly. “The oceans and their capacity for resilience are fi- nite,” stressed Professor Dr. Oliver Zielinski. The marine A camera mounted above the conveyor belt on the boat captures all the waste picked up during the clean-up mission. The image data is sent to a laptop in the cabin where it is evaluated using AI technologies. physicist is not only the head of Marine Perception at the DFKI Lab - Lower Saxony, but he is also scientific director of the AI Competence Center for Environment and Sustainability (DFKI4planet), which he founded in 2020. The United Nations published global sustain- ability goals as well as Germany‘s recently launched environmental offensive against marine pollution are particularly important, and projects such as AILHA should contribute directly to achieving these goals. “If we are able to apply AI effectively in the future to protect the environment and the climate, it can indeed contribute something towards saving the planet.” Prof. Dr. Oliver Zielinski, Head of research department Marine Perception, DFKI Niedersachsen More information www.dfki.de/dfki4planet www.berky.de www.everwave.de Contact Prof. Dr. Oliver Zielinski Head of research department Marine Perception DFKI Niedersachsen Head of DFKI4planet oliver.zielinski@dfki.de +49 441 99833 4710 The collection boat Moringa by Landmarken has operated since March taking waste out of the Mekong, Tonlé Sap, and Bassac Rivers in Phnom Penh. e v a w r e v e ©

BARRIERS BARRIERS 27 2 2 0 2 / 1 s w e N I K F D © pch.vector/Freepik practical utility as well as ease of operation. The design of the user interface was deliberately simplified to ensure in- tuitive operation. Follow-on project expected The app is to be further developed and prepared for practi- cal deployment in a follow-on project. Against the backg- round of the war in Ukraine and the large number of refu- gee children of daycare age, the functionality will also be expanded to include the language pairs Ukrainian/German, and Russian/German. Contact Dr.-Ing. Stefan Schaffer Research Department Cognitive Assistants stefan.schaffer@dfki.de +49 30 23895 1820 © FRÖBEL Bildung und Erziehung gemeinnützige GmbH The language barrier is part of everyday life at many educa- tional institutions, and it often makes it dicult to establish relationships between staff and parents. A good exchange between parents and educators, especially in early child- hood education, is essential in meeting the individual needs of children and strengthening them in their development. The aim of the SpeechTrans4Kita project is to break down language barriers between educators and parents with the aid of an application. Developed in a joint project bet- ween DFKI and FRÖBEL, the app is planned for initial use in FRÖBEL facilities. The technology used in conventional translation pro- grams is, for the most part, not mature enough to adequa- tely handle the translation of special pedagogical terms and the languages relevant to most daycare situations. Innovative language technologies An app developed in SpeechTrans4Kita is designed to ame- liorate the communication problems between parents and staff. The app currently provides both a written and an au- dible translation of Arabic and German so that communi- cation can take place quickly and easily. Expansion to ac- commodate other languages is an option. Through constant use and with the help of innovative lan- guage technologies and AI methods like machine learning, the app is trained and tailored for optimal use in early child- hood educational institutions. Data protection is also high-priority focus. Because privi- leged data is exchanged in communication between pa- rents and educational professionals, all data is processed and stored in accordance with the GDPR regulations. In October 2021, an initial demonstrator was tested on a group of test subjects and found to benefit both parents and educators. The app is characterized by a high level of

H b m G d r o N t i e h d n u s e G d n u b r e v s u a h n e k n a r K © The Bremen-based hospital group Gesundheit Nord and DFKI Cyber-Physical Systems research area created the “RIDIMP” project (Risk Indicators for cardiopulmonary Decompensation during Intensive care by Monitoring vi- tal Parameters) to advance this technology. What counts as a critical situation, and how do we assess it? The data at the intensive care unit must be analyzed before gaining any meaningful learning. The physicians at the Mid-Bre- men Clinic, which belongs to Gesundheit Nord, have de- fined two numerical assessment scores made up of many single parameters (like oxygen intake, breathing, pulse rate, or medication dosis) that, on the basis of patient data, show the state of circulation or respiration on a scale of 0 (non-critical) to 9 (highly critical). The scores are then subjected to machine learning techniques to assess ex- isting historical patient data and to execute a warning of the probability of a respiratory or circulatory collapse by predicting the future value of the score. Using the large volume of recorded data, the methods developed in the project provide very precise predictions about the evolution of the two scores and the future health of the patient. The project has achieved an error classifica- tion rate of 95 percent. This could alert the medical staff at the early stages of an impending problem. However, these are still early research results. The methods must be med- ically evaluated before they can be integrated into the or- ganization and procedures at the hospital to the benefit of the patients. The staff of the intensive care unit is confronted with a multi- tude of display screens, notifications, and alarm signals. Project RIDIMP is part of the overall KI-SIGS project, which has the aim of creating an “AI-Space for intelligent health- care systems.” Medical technology companies and part- ners at university clinics collaborate with AI institutes in northern Germany, specifically Bremen, Hamburg, and Schleswig-Holstein in several sub-projects to develop and build a dedicated AI platform. The German Federal Ministry of Economic Affairs and Climate Action (BMWK) sponsors KI-SIGS to promote better and faster development and implementation of medical AI technologies. More information www.ki-sigs.de Contact Prof. Dr. Christoph Lüth Research Department Cyber-Physical Systems christoph.lueth@dfki.de +49 421 218 59830 “The project demonstrates that AI platforms and applications are always the result of inter- action between subject matter experts – in this case, physicians and AI scientists. It takes the existing domain knowledge combined with the technical expertise of our AI researchers to develop powerful solutions. To me personally, the most satisfying thing about the medical application area is that at the end of the line, people directly benefit from our solutions.” Prof. Dr. Christoph Lüth, Deputy Head of DFKI Cyber-Physical Systems “Intensive care patients can quickly find themselves in a life-threatening situation associated with organ failure. An immediate response is demanded by medical experts. AI-aided systems support the early identification of critical scenarios and help to improve patient safety. The continuous improvement in treatment quality and patient safety is our primary goal.” Dr. med. Karin Hochbaum, MBA, Head of Business Development/Medical Strategy for the German hospital group Gesundheit Nord. 29 2 2 0 2 / 1 s w e N I K F D d r o N t i e h d n u s e G / e s a H n i t s r e K ©

Dr. Shoya Ishimaru Named to DFKI‘s First Endowed Professorship at TU Kaiserslautern Prof. Andreas Dengel Honored with Prestigious Japanese Order of Merit in Artificial Intelligence 31 2 2 0 2 / 1 s w e N I K F D Dr. Shoya Ishimaru personally coined the term “Psyber- netics” to describe a field of research that is developing a sensing-feedback loop between humans, machines, and science. His goal is to use eye-tracking and physi- ological sensing systems to extend the measurable and controllable domain of cybernetics. This involves not only information space and physical space but also in- cludes psychological space, i.e., emotions, affects, and knowledge states. Ishimaru has long been interested in enhancing human cognitive abilities such as learning, thinking, and communicating through the development of new technologies. A longtime employee of the Smart Data & Knowledge Services department of DFKI, he earned his Ph.D. in 2019. That same year, he was awarded the title of MITOU Su- per Creator, which is only awarded to outstanding young researchers by the Japanese Ministry of Economy, Trade, and Industry. Now, the Junior Professor heads the inter- disciplinary research unit “Psybernetics Lab” of the Com- puter Science department at Kaiserslautern University of Technology (TUK). Ishimaru remains connected to DFKI through the Immersive Quantified Learning Lab (iQL). More information https://psyberlab.de Japanese Consul General Shinichi Asazuma present- ed Professor Dr. Andreas Dengel with the high ranking “Order of the Rising Sun, with Gold Rays and Neck Ribbon” in the name of His Majesty Kaiser Naruhito. The Order was established in 1875 and is awarded in various degrees. The design features rays of sunlight from the rising sun symbolizing the Japanese national flag “hi no maru.” The presentation ceremony was held at the Japanese Consu- late in Frankfurt am Main on April 13, 2022. The country‘s oldest award recognizes Den- gel‘s outstanding contribution to the academic exchange between Japan and Germany in the field of artificial intelligence and his contributions to promoting understanding between the two countries. Andreas Dengel is Executive Director and head of DFKI‘s Smart Data & Knowledge Services research department at DFKI-Kaiserslautern. He is also the Chair of the Artifi- cial Intelligence Department at Kaiserslautern Universi- ty of Technology and an AI Ambassador for the state of Rhineland-Palatinate. In Japan, he is a sought-after con- sultant for science, industry, and politics. He has initiated numerous collaborative projects in research and industry in the field of artificial intelligence and has long been an advocate of German-Japanese cooperation in the eco- nomic exploitation of artificial intelligence. Dr. Shoya Ishimaru. Prof. Andreas Dengel and Consul General Shinichi Asazuma.

Prof. Drechsler Named “Distinguished Member” of the Association for Computing Machinery (ACM) Prof. Dr. Rolf Drechsler, head of DFKI‘s Cyber-Physical Systems Department and the Computer Architecture WG at the University of Bremen, was named a Distinguished Member of the Association for Computing Machinery (ACM) on December 15, 2021. The honor came in recognition of his outstanding achievements in the field of computer science and his many years of mem- bership in the ACM. Photo: Lisa Jungmann / University of Bremen 33 F E I R B N I S W E N 2 2 0 2 / 1 s w e N I K F D M C A : o g o L © The nominees for this award must have at least 15 years of professional experience in the field and five years of ACM mem- bership in the past ten years, as well as significant achievement or influence in the field of computer science, informatics, or information technology. A Distinguished Member of the ACM is expected to act as a mentor and role model, driving techni- cal developments and making an above-average contribution to this special research field. More information: www.acm.org/media-center/2021/december/distinguished-members-2021 DAAD Postdoc-NeT-AI: Robotics Innovation Center Hosts International Young AI Talents The Postdoctoral Networking Tour in Artificial Intelligence (Postdoc-NeT-AI) of the German Academic Exchange Ser- vice (DAAD) offers outstanding young researchers from all over the world an opportunity to network with leading AI scientists in Germany. The DFKI Robotics Innovation Center (RIC) participated as a host institution with a focus on robotics in this year‘s Postdoc-NeT-AI and introduced itself to the participants during the virtual networking week from March 7 to 11, 2022. In the run-up phase, researchers from the RIC selected in- terested candidates and invited them to virtual one-on- one meetings. In addition, DAAD enabled participants to take one-week individual trips to visit institutions and in- dividual researchers. Contact: Dr. Sirko Straube, sirko.straube@dfki.de Two New German-French Collaborative Research Projects In 2022, DFKI and the French computer science institu- te Inria have launched two new projects, ENGAGE and R4Agri, as part of their strategic research and in- novation agenda in the field of Artificial Intelligence. On January 22, 2020 – the first anniversary of the signing of the Aachen Treaty – the two research organizations signed a Memorandum of Understanding agreeing to significantly expand their collaboration in the field of AI and to structure and formalize their long-standing scientific cooperation. ENGAGE investigates how to optimize high-performance com- puting (HPC) environments, making them more ecient for AI when used in concert with other hardware environments. In R4Agri, a project team is developing an AI-based recom- mendation engine to support decision-making. This basic research has potential for application, for example, in the area of agricultural robotics. In parallel, the findings from these projects will flow into DFKI‘s transfer projects for the digital transformation in the agricultural sector.

Intelligent Solutions for the Knowledge Society Bremen Site Prof. Dr. Rolf Drechsler Cyber-Physical Systems Prof. Dr. Dr. h.c. Frank Kirchner Robotics Innovation Center Oldenburg Bremen Osnabrück Branch Office Lübeck Prof. Dr. Heinz Handels AI in Medical Imaging Prof. Dr. Alfred Mertins AI in Biomedical Signal Processing Prof. Dr. Ralf Möller Stochastic Relational AI in Healthcare Lübeck Laboratory Niedersachsen Osnabrück Prof. Dr. Joachim Hertzberg Plan-Based Robot Control Prof. Dr. Oliver Thomas Smart Enterprise Engineering Berlin Laboratory Berlin Prof. Dr. Gesche Joost Interactive Textiles Prof. Dr. Volker Markl Intelligent Analytics for Massive Data Prof. Dr. Sebastian Möller Speech and Language Technology Prof. Dr. Niels Pinkwart Educational Technology Lab Oldenburg Prof. Dr. Daniel Sonntag Interactive Machine Learning Prof. Dr. Oliver Zielinski Marine Perception Kaiserslautern Saarbrücken Saarbrücken Site Prof. Dr. Josef van Genabith Multilinguality and Language Technology Prof. Dr. Antonio Krüger Cognitive Assistants Prof. Dr. Peter Loos Institute for Information Systems Prof. Dr. Wolfgang Maaß Smart Service Engineering Prof. Dr. Philipp Slusallek Agents and Simulated Reality Kaiserslautern Site Prof. Dr. Prof. h.c. Andreas Dengel Smart Data & Knowledge Services Prof. Dr. Paul Lukowicz Embedded Intelligence Prof. Dr. Martin Ruskowski Innovative Factory Systems Prof. Dr. Hans Schotten Intelligent Networks Prof. Dr. Didier Stricker Augmented Vision Prof. Dr. Sebastian Vollmer Data Science and its Applications k k i i p p e e e r r FF ©©©© y y z e e ee e e e t t c c c c c e e e e e e v v v / / / / s s s s t t t t r r r r a y r r a H © Living Labs Testing, evaluation, and demonstration of innovative technologies in comprehensive application scenarios Advanced Driver Assistance Systems Living Lab, Bremen Ambient Assisted Living Lab, Immersive Quantified Learning Lab (IQL), Innovative Retail Lab, Robotics Exploration Lab, Smart City Living Lab, SmartFactory-KL, Smart Office Space Living Lab Competence Centers Coordination of research activities in particular areas AI for Environment and Sustainability, Ambient Assisted Living, Autonomous Driving, Deep Learning, Emergency Response and Recovery Management, Safe and Secure Systems, Semantic Web, Smart Agriculture Technologies, Wearable AI