1. INRIA-DFKI EUROPEAN SUMMER SCHOOL ON ARTIFICIAL INTELLIGENCE Trustworthy AI • AI for Medicine Trustworthy AI AI for Medicine SACLAY, FRANCE JULY 20-23, 2021 IDAI 2021 IDAI 2021 is the first joint Summer School of Inria (Institut national de recherche en informatique et en automatique) and DFKI. IDAI 2021 is designed for PhD students, MSc students, postdocs, and researchers in industry in all areas of AI, including: Registration deadline: April 19, 2021 https://idessai.inria.fr IDAI 2021 will be held as an in-person or online event, depending on the Covid 19 infection situation. • Machine learning • Knowledge representation and reasoning • Search and optimization • Planning and scheduling • Multi-agent systems • Language technology • Robotics • Computer vision
03 L A I R O T I D E 1 2 0 2 / 1 S W E N I K F D INDUSTRIAL AI, NATIONAL INITIATIVES, AND INTERNATIONAL COOPERATION The first article to use the term INDUSTRIE 4.0 was co-authored by Prof. Kagermann, Prof. Lukas, and Prof. Wahlster and was published on April 1, 2011. The concept spread rapidly and was taken up by major industry players and by small and medium-sized enterprises. New, flexible "Smart Factories" have been built and existing plants have been retrofitted to make them compatible with INDUSTRIE 4.0 concepts. Essential in all of this was the application of industrial AI. The semantic product memory, the virtual representation of products and manufacturing environments – keyword: digital twins – ensure that the production of even the smallest lots is economically viable (p. 6). In the second wave of the 4th Industrial Revolution, AI systems use digital data to create new value chains and innovative business models. Flexible manufacturing systems are moving forward with hybrid teams and the advances in collaboration among people and machines (p. 16). DFKI presents projects, software, and research prototypes to demonstrate how this all works at the Hannover Messe Digital Edition 2021 (p. 12). The ray-tracing system, which originated at Saarland University and further developed at DFKI, won the "Technology Oscar" from the Academy of Motion Picture Arts and Sciences. Based on simulated beams of light, the system makes it possible to correctly depict surfaces with all their shades and re- flections (p. 35). DFKI continues to expand in northern Germany with a focus on research on AI in medicine. The state of Schleswig-Holstein established a branch office with three new research departments in the city of Lübeck, and provided a funding amount of three million euros (p. 20). Medicine is also one of the fields of application being studied at a new research unit in the city of Oldenburg. The main subject of re- search is the interaction of people with AI systems (p. 18). Additional initiatives and collaborations have emerged on a national and international level. The Eu- ropean Space Agency ESA and DFKI have founded a joint research laboratory at DFKI-Kaiserslautern with the aim of developing new AI technologies and applications for use in commercial space travel (p. 27). The Federal Office of Criminal Investigation (BKA), Rhineland-Palatinate's state Office of Crim- inal Investigation (LKA), and DFKI have launched a transfer lab to develop AI technologies for use in police matters (p. 30). DFKI is also making plans to set up a branch in Japan. Osaka Prefecture University (OPU), which has links with DFKI, has agreed to the procedure for opening a DFKI research laboratory (p. 29). Another suc- cess in the internationalization of INDUSTRIE 4.0 in Japan is the founding of the Flexible Factory Partner Alliance – FFPA – organized with the aim of promoting standardization of the coordination and control systems for wireless technologies in manufacturing (p. 28). Prof. Dr. Antonio Krüger Prof. Dr. Antonio Krüger Prof. Dr. Prof. h.c. Andreas Dengel Prof. Dr. Dr. h.c. Frank Kirchner Prof. Dr. Dr. h.c. Frank Kirchner Prof. Dr. Philipp Slusallek
4 0 CONTENT DFK I NE WS 1/2021 2226
05 T N E T N O C 1 2 0 2 / 1 S W E N I K F D 34 Innovative Learning Environment for Psychomotor Skills – Project MILKI-PSY Started 35 Prestigious „Sci-Tech Oscar“ Goes to Three Former Computer Science PhD Students from Saarbrücken 36 New Technologies to Support Post-Editing of Machine Translations Imprint Issue 47, April 2021, 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 email@example.com Phone +49 681 85775 5390 Translation team Glenn Peach, Armindo Ribeiro, Sylvia Krüger, Heike Leonhard Photos 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 37 CEF AT Tools and Services – Translation Tools from and for Europe 38 Transfer of Control Among Autonomous Agents – TRACTAT Project Ends Successfully 39 DFKI Staff Interview with Dr. Tim Schwartz 40 News in brief 42 Company Profile 31Content
06 D F K I N E W S 1 / 2 0 2 1 TEN YEARS OF INDUSTRIE 4.0 GERMANY DRIVING INDUSTRIAL AI AS THE MEANS TO FUTURE VALUE CREATION Almost ten years ago to the day, the article that initially introduced the engaging concept of INDUSTRIE 4.0 appeared. “INDUSTRIE 4.0: The 4th Industrial Revolution and the Internet of Things” was published by VDI (Association of German Engineers) on April 1, 2011, just in advance of the Hannover Fair (Messe) 2011. The article was co-authored by Prof. Wolfgang Wahlster, DFKI, Prof. Henning Kagermann, acatech, and Prof. Wolf-Dieter Lukas, Federal Ministry of Education and Research (BMBF). DFKI corporate spokesman Reinhard Karger caught up to Prof. Wolfgang Wahlster, founding director of DFKI and former CEO, who still accompanies the work of DFKI as Chief Executive Advisor (CEA), for this interview. Prof. Henning Kagermann, Prof. Wolfgang Wahlster and Prof. Wolf-Dieter Lukas.
08 D F K I N E W S 1 / 2 0 2 1 How did the term continue to spread even further? The concept rapidly spread virally around the world and is in use today, like the words “kindergarten” and “autobahn” with their association to Germany. INDUS- TRIE 4.0 is accepted as a German export hit that has received worldwide attention and recognition in busi- ness, science, and politics. called digital twin to each factory module and each unfinished product. A digital twin stores the function and the entire life his- tory of a physical object in a machine-understandable representation and actively controls the production pro- cess itself, making it economically feasible to produce even in very small lots. We knew what we were doing when we introduced the term “Industrie 4.0” to the world. This is the first time we have been able to establish an innovative concept from Germany in the international world of high tech, where for so many years the terms have originated in America or Asia. In the last ten years, more than 100,000 publications, 10,000 conferences, and 1,000 project consortia have addressed the techni- cal implementation of INDUSTRIE 4.0; resulting in more than 25 million Internet references to the term. Did any preliminary work exist on which to base INDUSTRIE 4.0? Yes, for the specification of INDUSTRIE 4.0 objectives, we built on the research results of our world’s first “Smart Factory” at DFKI-Kaiserslautern, which has been under development since 2005. Also, I had been leading the BMBF SemProm project on semantic product memo- ries since 2008, and, based on the demonstrator op- erated by colleague Detlef Zühlke, we assigned a so- How many of the “INDUSTRIE 4.0” con- cepts that you and your colleagues Ka- germann and Lukas published on April 1, 2011 have since become today’s reality? The Internet of Things and the cyber-physical systems that use it already exist in modern factories today. The digital connectivity between all machines, tools, work- pieces, and factory workers has also made tremendous progress in existing factories. Steady advances in new low-cost sensors and wireless links enable the upgrade of the sensor systems so more and more production steps can be monitored by multi-sensor fusion in real-time, for example, for quality control and other purposes. More and more digital twins are also being realized on the basis of our earlier work on active digital object memories. We are nearing the goal of having a digital image that can be automatically updated in real-time Industrial AI enables direct cooperation between humans and machines.
10 D F K I N E W S 1 / 2 0 2 1 What does international cooperation on INDUSTRIE 4.0 look like from a DFKI perspective? The state visit by Chancellor Angela Merkel to the Czech Republic in August 2016 provided the framework for the launch of a bi-national initiative with former Czech Prime Minister Bohuslav Sobotka, which resulted in the establish- ment of the German-Czech Innovation Lab for Human-Ro- bot-Collaboration in INDUSTRIE 4.0 located in Prague. The cooperation with colleagues at the Czech Technical Uni- versity in Prague is excellent. This Research and Innova- tion Center for Advanced Industrial Production (RICAIP) will present its findings at the digital Hannover Fair 2021. The research focus is on Human-Robot-Collaboration in industrial manufacturing. The goal is to develop hybrid teams of humans and robots working together, preferably side by side. For this to work, machines need to know the context of an action and the goal, so they can interpret the status and evaluate their contribution to a solution. We also have Smart Factory cooperation agreements with France and the Netherlands. China is currently making very targeted investments in order to use INDUSTRIE 4.0 to transform its manufacturing sector, which just a few years ago was still at an Industry 2.0 level, to be the best in the world. Japan is already further along because they understood the new paradigm faster and more deeply, and Korea is also catching up. The USA was late to rec- ognize the full scope and depth of the paradigm shift and got off to a false start with the initial one-sided focus on the “Industrial Internet.” Fundamental work on INDUSTRIE 4.0 was not started there until much later. What new technology trends will impact the 4th Industrial Revolution in the next ten years? Six new megatrends will be crucial. These are: industri- al AI, edge-computing (to include the edge cloud), 5G in the factory, team robotics, autonomous intra-logistics systems, and reliable data infrastructures like the one provided by GAIA-X. Industrial AI will power the second wave in digital manu- facturing. The first wave, the digitalization of all produc- tion and supply chain data and its availability via cloud systems, is now largely complete. Today, this digital data can be analyzed by AI systems and interpreted in context, enabling its active use in the future in new value chains and innovative business models. Presentation of initial recommended actions at Lab Talks during CeBIT 2012, 6. March 2012. „INDUSTRIE 4.0: Cyber-physical systems as the basis for the 4th Industrial Revolution. (l.to r.) Prof. Dr. Lutz Heuser, AGT Group Germany; Dr. Andreas Goerdeler, Federal Ministry of Economics and Technology, BMWi; Prof. Dr. Willem Jonker, EIT ICT Labs; Prof. Dr. Wolf-Dieter Lukas, Federal Ministry of Education and Research, BMBF; Prof. Dr. Wolfgang Wahlster, DFKI; Prof. Dr. Henning Kagermann, acatech.
12 D F K I N E W S 1 / 2 0 2 1 / ONLINE / DFKI AT HANNOVER MESSE 2021 DIGITAL EDITION Hybrid robotic teams in dynamic industrial environments, generation and exploitation of data products through smart services, comprehensive perception of complex situations in environmental, agricultural, and business processes, AI in medicine, for logistics, and trade – from Monday, April 12, to Friday, April 16, 2021, at the HANNOVER MESSE 2021, the German Research Center for Artificial Intelligence will present numerous innovations, project results, and research prototypes for the entire entrepreneurial value chain – 100% digital. The world's leading trade fair for industry is all about "Industrial Transformation". This year, the DFKI is looking back on ten years of INDUSTRIE 4.0. The industrial transformation is progressing relentlessly and is being driven by the three megatrends of digitization, individualization, and climate protection. DFKI and its project partners address related issues with their application-oriented research and present industry-oriented results from the AI research fields of deep learning, robotics, sensor technology, human-robot collaboration, language and education technology, GreenTech, smart services in industrial and business processes, Gaia-X, explainabil- ity, and trustworthiness of AI systems. All information about the exhibits, live streams and events, all videos and flyers: www.dfki.de/hm21 Contact Christof Burgard Head of Corporate Communications DFKI Saarbrücken firstname.lastname@example.org +49 681 85775 5277 Heike Leonhard Corporate Communications DFKI Saarbrücken email@example.com +49 681 85775 5390
13 1 2 0 2 / 1 S W E N I K F D AdEPT BIONIC Augmented Reality and AI-based Learning, Teaching, and Collaborating Intelligent Sensor Networks Reduce Physical Stress CHIM CLAIRE Innovation Network Chatbots at the Museum AI Made in Europe CoMem AI to Assist Daily Work in Companies Cyber-Physical Systems Verification and Virtual Prototyping for RISC-V-Systems Franco-German Cooperation in AI DFKI4planet Inria (Institut national de recherche en informatique et en automatique) and DFKI AI for Environment and Sustainability European Language Grid (ELG) EVAREST Scalable Cloud Platform for Multilingual Europe Development and Utilization of Data Products in Food Industry IIP-Extrem InGewA Individualized Implants and Prosthetics for the Lower Extremities The Integrated Trade Tax Wizard (available in German only) Innovative Retail Laboratory (IRL) KI-Campus Application-Oriented Research for the Retail of the Future The Learning Platform for Artificial Intelligence Production Level 4-Demonstrator Robotics Innovation Center The World's First Production Level 4 Demonstrator Intelligent Robotics - for the Earth, the Space and the Human Smart Construction SPAICER Artificial Intelligence in Construction Smart Resilience Services in the Manufacturing Industry TexaS TRACTAT, CAMELOT, RICAIP Construction Kit for Multifunctional Textile-adapted Electronic Systems Transfer of Control in Distributed INDUSTRIE 4.0 Use Cases Xaines Explaining AI with Narratives
16 D F K I N E W S 1 / 2 0 2 1 Various collaborative robot systems are studied in the HRC lab at the Robotics Innovation Center in Bremen. A SAFE TEAM: THE ROBOTICS INNOVATION CENTER WORKING ON FLEXIBLE COOPERATION BETWEEN HUMANS AND ROBOTS Picture the massive robotic arms on an assembly line that assembles cars in large numbers at high speed. This is the image that probably comes to mind when you hear the words “industrial robotics.” Automated production is the most widely known area of use for stationary robot- ic arms. These heavy industry robots keep to themselves for safety reasons and operate in segregated areas to car- ry out pre-programmed steps. Researchers in the field of human-robot collaboration, however, are working on how to exploit the potential of hybrid teams. After all, numerous advantages can ensue when the helping hands are equipped with modern sensor systems and Artificial Intelligence (AI). Collaborative robot systems can be a great help, for ex- ample, in smaller manufacturing companies: They provide “The vision of human-robot collaboration is that people will be able to perform their jobs with no change to the workplace and bring in the robots when needed.” Dennis Mronga, Head of the research field Collaborative Robotics at the DFKI Robotics Innovation Center (RIC) flexible support in various work steps, require little space, carry bulky loads, and can tighten screws in places that are difficult for humans to reach – not in a secured industrial plant, but in a commonly shared workplace. “The vision of human-robot collaboration is that people will be able to perform their jobs with no change to the workplace and bring in the robots when needed,” said Dennis Mronga, head of the research field Collaborative Robotics at the DFKI Robotics Innovation Center (RIC). Robot systems that can work alongside humans still make up only five percent of the international market for ro- bots. This is about to change: according to a study by the consulting firm ABI Research, collaborative robotics will become increasingly mainstream in the next decade. The study was released in 2019 when about $700 million was spent on collaborative robot arms. The firm forecasts annual sales of nearly $12 billion by 2030 – just under 30 percent of the entire robotics market. Collaborative robot systems must react safely to the motions of tools like gripper and puller arms.
However, there is still a long way to go before this pro- jected revenue is reached. The operating speed and cost of the collaborative robot arms are major factors. But several questions stand out above all else: How do we guarantee that humans and robots can interact safely? How can we avoid crush injuries, impact injuries, and haz- ards from sharp-edged materials? The research team of Dennis Mronga is working on the answers to these ques- tions at the Human-Robot-Collaboration (HRC) Labs of DFKI-Bremen. One obvious consideration for safe interaction is the de- sign of the robot: An angular design increases the risk of injury and a heavyweight may lead to crushing and im- pact injury. Ideally, a rounded and lightweight robot arm can be designed. But, the role played by humans must also be considered: adequate training of staff is impor- tant to counter the fear and uncertainty in dealing with a robot colleague if we are to ensure safe cooperation in a hybrid team. Robot arms will still be equipped with programmed com- pliance, which ensures the system can react to external influences. Similarly, the arm will interrupt its motion if it encounters resistance – something like a human hand. But, a collaborative robot can react to humans even without physical contact: motion sensors allow it to avoid colli- sions and to reliably detect humans even from a distance. “Probably the safest solution would be a skin of proximity sensors for the robot arm, which provides absolute con- tact avoidance – but, in many situations, the human must be able to control the arm,” explained Dennis Mronga. 17 1 2 0 2 / 1 S W E N I K F D brain waves can detect fatigue or overwork, for exam- ple, triggering the implementation of an increased level of safety. The method is to be released as a cloud-based service for cognitive occupational safety. However, a few more steps are necessary before the wide use of robots in hybrid teams becomes a reality: Un- like segregated assembly-line robots that are equipped with a few sensors and can only adapt their activities in a limited way, the collaborative systems require a lot of environmental information, which is processed at high speeds and incorporated in their planned movements. And, unlike in the classic “Sense-Plan-Act-Methodology,” the robot has to adapt in real-time to operate alongside humans in a dynamic environment. The two-armed KUKA system at the HRC Lab is used in assembly scenarios such as in handling tools. While the safety issue affects most systems and manu- facturing plants, the collaborative robots of the future will have an attribute that promises a key advantage when working in hybrid teams, i.e., the ability to respond in- telligently to their surroundings and make autonomous decisions. Artificial intelligence enables control meth- ods that intuitively respond to humans and reduce risk through real-time processing of sensor data. Contrary to common beliefs, autonomous systems can be taught not to act against humans but rather to actively protect them by moving away. Several AI technologies already exist for this purpose. Nevertheless, hybrid teams make sense if the complex- ity of the tasks is already manageable for the robot: For example, if instead of a gripper arm – which is one of the most complicated actions for robot arms – a screw- ing function is sufficient, more capacity can be used for safety. The safer and more flexible the systems become, the greater the potential away from the assembly lines – whether as a spontaneous helper in the small business, or for ergonomic relief with a contorted assembly. “A robot arm does not need to be able to do everything before it starts to relieve people ergonomically or to support them with monotonous tasks,” said Dennis Mronga. In addition to considerate movements, collaborative ro- bots can use other means to increase safety in the hybrid team. For example, light signals indicate how or where the system will move next. But things are more complex for the research team of Dr. Elsa Kirchner: Project KAMeri aims to measure the cognitive state of the human using an EEG headset and then use this data to adjust the be- havior of the collaborating robotic systems. The captured More information www.dfki.de/ric Contact Dennis Mronga Research Deptartment Robotics Innovation Center firstname.lastname@example.org +49 421 17845 6560
AI FOR SPACEFLIGHT – AI FOR SPACEFLIGHT – ESA AND DFKI LAUNCH ESA AND DFKI LAUNCH JOINT TRANSFER LAB JOINT TRANSFER LAB 27 1 2 0 2 / 1 S W E N I K F D ESA's OPS-SAT, a mini satellite launched in 2019, allows European industry and academic experts to test innovative new software that could facilitate pattern recognition, autonomous scheduling, deep learning and automated maneuvering in orbit. The European Space Agency (ESA) and the German Research Center for Artificial Intelligence (DFKI) are establishing a joint research laboratory to develop new AI technologies and systems for use in commercial spaceflight – the ESA_Lab@DFKI. The Transfer Lab at DFKI-Kaiserslautern creates a framework for scientists from both research organizations to study AI systems for use in the interpretation of complex, extensive earth observation data and collision avoidance systems for satellites. “The potential for beneficial AI applications in space in the face of a world undergoing technological and climatic change is huge. In this regard, Machine and Deep Learning methods, for example, are ideal for analyzing and interpreting the extensive and complex data recorded by Earth observation systems, whether for climate monitoring, disaster control, or agriculture. The new transfer lab provides a creative space for the joint development of tailored space applications and the identification of sustainable cutting-edge and future needs AI solutions in ESA projects.” Prof. Andreas Dengel, Executive Director at DFKI in Kaiserslautern and head of the Smart Data and Knowledge Services research department Artificial intelligence for Earth observation Satellites send large amounts of data back to Earth every day. AI methods can help generate valuable knowledge from this raw data. Special machine learning methods are used, for example, to make dispersion and dam- age forecasts for environmental protection and disas- ter management in areas affected by natural disasters. Commercial value-added services range from financial risk assessment of such events to monitoring industrial infrastructures on Earth. Another application scenario targets security of supply through agricultural products: yield forecasts are made after the analysis of growth conditions and soil qualities of cultivated areas. Artificial intelligence for the safe operation of satellites Another potential field of application for AI is collision avoidance systems for spacecraft threatened by more and more orbiting objects above Earth. In the past, large, single, highly specialized satellites were in operation. Today, the trend in international spaceflight is toward constellations of hundreds or even thousands of small satellites. This poses new challenges for the operators of space-based infrastructure because, as the number of satellites increases, so does the risk of collision and the creation of even more space debris. AI methods accurately calculate the orbits of active and passive satellites as well as pieces of space debris, helping to prevent collisions by initiating evasive maneuvers. Successful transfer from the research ecosystem The partnership between ESA and DFKI in form of the ESA_Lab@DFKI will support these and other basic research developments and promises to expand the range and scope of innovations transferred from academic research to advanced industrial applications. More information www.dfki.de/sds www.esa.int/Space_in_Member_States/Germany Contact Dr. Mateus Dias Ribeiro Research Department Smart Data & Knowledge Services email@example.com +49 631 20575 1017
28 D F K I N E W S 1 / 2 0 2 1 WIRELESS COMMUNICATION IN FACTORY ENVIRONMENTS – FFPA DRIVING CONNECTIVITY FOR FFPA DRIVING CONNECTIVITY FOR MANUFACTURING FACILITIES MANUFACTURING FACILITIES A high degree of flexibility and productivity are key requirements for modern produc- tion in INDUSTRIE 4.0. Wireless communication is essential for this. The FFPA – Flexible Factory Partner Alliance – is a non-profit organization founded in Japan, promoting the standardization of coordination and control technology for wireless technologies in production. It is chaired by Prof. Andreas Dengel, executive director DFKI Kaiserslautern and Prof. Hans Dieter Schotten, head of DFKI Research Department Intelligent Networks. “Flexible Factory” represents an evolved site for flexible on-demand manufacturing of variable product types with variable production volumes, emphasizing mobility and configurability of manufacturing facilities. Wireless connectivity plays an important role for factory networks accommodating frequently relocated production ma- chines and equipment and increasing retrofitted sensors. In such factories, wireless systems provided by different vendors often coexist. They use wireless technologies in different generations and with different standards, re- sulting in communication problems due to unexpected mutual interference. The Flexible Factory Partner Alliance (FFPA) has been es- tablished in 2017 as a non-profit organization to promote the formulation of standards for coordination control technology, ensuring stable communications in an en- vironment where various wireless systems coexist in the manufacturing facilities. FFPA supports the SRF (Smart Resource Flow) wireless platform that enables coordination control among various wireless systems for stable operation. On this platform, visualization and integrated management for wireless networks and the associated devices and systems are easy to implement. FFPA has already defined the tech- nical specifications Version 1.1 and the conformance test specifications. For 2021, the Alliance is planning a pro- gram to certify products that are complying with the SRF wireless platform. As the next step, FFPA has been developing the technical specifications Version 2.0 mainly to support 5G, the fifth generation of mobile communication. 5G is expected to reinforce the SRF wireless platform’s functionality by its high performance in broadband, low-latency and mas- sive connectivity. In this context, FFPA signed a Memo- randum of Understanding with 5G-ACIA (The 5G Alliance for Connected Industries and Automation) in April 2020 to cooperate and build a global ecosystem for industrial wireless communications, particularly regarding indus- trial 5G. Starting from resolving wireless communication problems in factories, FFPA is continuously contributing to the realization of technical and business platforms for enhancing productivity of manufacturing by using ICT, including IoT and AI. More information www.ffp-a.org Contact Andreas Dengel Chair Flexible Factory Partner Alliance firstname.lastname@example.org
34 D F K I N E W S 1 / 2 0 2 1 INNOVATIVE LEARNING INNOVATIVE LEARNING ENVIRONMENT FOR ENVIRONMENT FOR PSYCHOMOTOR SKILLS PSYCHOMOTOR SKILLS PROJECT MILKI-PSY STARTED PROJECT MILKI-PSY STARTED A scientific presentation on YouTube, TED Talks, digital lectures, and seminars – learning even abstract know- ledge and complex issues no longer requires a physical presence. Teaching psychomotor skills in a university ed- ucation today, however, still does. The motion sequences for a sport, or a surgical operation, or manual dexterity in the arts and crafts – all depend on practical exercise, direct feedback, and reflection. Learning success relies on personal support and material input, which limits the scalability of such study courses. This may change with current technological developments promising new opportunities to convey motor knowledge via distance learning. For example, immersive learning and practice rooms in mixed, augmented, and virtual re- ality settings or the latest sensor systems that can track and record movements at a highly detailed level already exist. Big Data and learning analytics now analyze and evaluate large volumes of data. Machine learning and generative AI models interpret this data, draw conclu- sions, and generate individual feedback. Each of these technologies has been treated separately – until now. On the basis of complex data analyses, AI-based analyt- ics facilitate learning progress through automated error detection and personalized feedback. MILKI-PSY is one of twelve new projects funded by Ger- man government research grants dedicated to innova- tion through artificial intelligence and big data in the field of “Digital Higher Education.” PROJECT SUMMARY Consortium German Sport University Cologne Sponsor Volume Term DFKI Leibniz Institute for Research and Information in Education RWTH Aachen University Cologne University of Applied Sciences (lead management) Federal Ministry of Education and Research (BMBF) 2.3 million euros March 1, 2021 - February 29, 2024 More information www.dfki.de/edtec https://wihofo.bmbfcluster.de/de/milky-psy-3775.php Contact Dr. Milos Kravcik Research Department Educational Technology Lab email@example.com +49 30 23895 5513 The aim of the MILKI-PSY (Multimodal Immersive Learn- ing with Artificial Intelligence for Psychomotor Skills) re- search project is to integrate these technologies to cre- ate an independent, immersive learning environment for psychomotor skills using AI-supported, data-intensive, multimodal methods. This comprehensive approach will enable multimodal recordings of the experts’ activities and the use of these recordings as blueprints for learners.
35 1 2 0 2 / 1 S W E N I K F D PROJEKT XAINES ZUR PRESTIGIOUS “SCI-TECH OSCAR” GOES ERKLÄRBARKEIT VON KI GESTARTET TO THREE FORMER COMPUTER SCIENCE PHD STUDENTS FROM SAARBRÜCKEN The Academy of Motion Picture Arts and Sciences awards the distinction known as the “Oscar” also for technical merit. 2021 the basic research and technological development of the “Intel Embree Ray Tracing Library” will be honored with the Technical Achievement Award. Three award winners earned their doctorates under Philipp Slusallek at Saarland University. The university professor and DFKI researcher has significantly advanced the award-winning ray tracing tech- nology over the past two decades. Today, this technology provides many Hollywood movies and computer games with the perfect images and can also recently be found in almost all graphics processors. At the virtual award ceremony on February 13, Sven Woop, Carsten Benthin, and Ingo Wald received the Sci-Tech Oscar. All three earned their doctorates more than a decade ago at Saarland University on the topic of ray tracing. Together with Attila T. Áfra and Manfred Ernst, they are honored for co-developing or doing preparatory research on the “Intel Embree Ray Tracing Library.” This software library was de- veloped by Intel mainly for its own processors (CPUs). It is also freely available worldwide to all film producers and other applications as an open-source platform. In particular, the software library has been used to make animated films and virtual game worlds look as realistic as possible. It has been employed, for example, in international hit movies such as Lego Batman, Spider Man, and The Grinch, as well as in the computer game Cyberpunk 2077. “In animated films or virtual role-playing games, it is im- portant to correctly depict surfaces with all their shades and reflections. In each scene, it is also necessary to recal- culate which objects are visible to the viewer and how the incidence of light changes during a movement,” explains Philipp Slusallek. As a computer graphics professor at Saar University, Philipp Slusallek and his team were instrumental in developing the basic ray tracing technology. L.t.r. Ingo Wald, Carsten Benthin and Sven Woop. As early as 2003, the current Oscar winners, together with Professor Slusallek, founded the company “inTrace,” which initially marketed the ray tracing process for the automotive industry. The car manufacturers could use it to simulate the different variants of a vehicle’s body or interior in a photo- realistic way. The high computing power that the ray trac- ing process demands of computers attracted the attention of chip manufacturer Intel to the Saarbrücken researchers. In 2009, the company funded the establishment of an Intel Visual Computing Institute at Saarland University with 12 million dollars - and recruited the three young researchers Carsten Benthin, Sven Woop, and Ingo Wald. Philipp Slusallek now uses the method for a wide range of applications at the DFKI in Saarbrücken, where he is now also managing director. “Among other things, we use real-time ray tracing today to generate synthetic sensor data very quickly, which can be used to train and validate AI systems. The technique is also used for radar simulation, which plays an important role in autonomous driving. In addition, day- lighting systems in architecture can be optimized with it,” explains Professor Slusallek. „I think you can already say that we have advanced the computer world a little bit from Saarbrücken.“ Prof. Dr.-Ing. Philipp Slusallek More information www.oscars.org/sci-tech/ceremonies https://graphics.cg.uni-saarland.de www.dfki.de/asr Contact Prof. Dr.-Ing. Philipp Slusallek Head of Research Department Agents and Simulated Reality firstname.lastname@example.org +49 681 85775 5377
36 D F K I N E W S 1 / 2 0 2 1 NEW TECHNOLOGIES NEW TECHNOLOGIES TO SUPPORT POST-EDITING TO SUPPORT POST-EDITING OF MACHINE TRANSLATIONS OF MACHINE TRANSLATIONS The quality of machine translation has improved significantly in recent years. Translators are increasingly shifting their activities to post-editing of machine translations. This saves time, reduces errors, but changes the way they inter- act with the text. In the MMPE project (Multi-modal and Language Technolo- gy-based Post-Editing Support for Machine Translation), an interdisciplinary team of DFKI researchers investigated how post-editing can be technologi- cally supported. Post-editing (PE) combines the advantages of artificial in- telligence and human intelligence; it also shifts the focus of translation work: Instead of generating text, transla- tors correct errors in otherwise helpful suggestions in the target language. Improving the frequently recurring ma- chine translation (MT) errors is tedious; fixing hard-to-find or complex errors makes the job cognitively demanding. To avoid repetitive errors of MT, the researchers have in- vestigated various deep-learning architectures for au- tomatic post-editing (APE) that can adapt the output of each black-box MT system to a particular domain or style. Rather than learning to translate, APE systems learn from recurring human corrections and apply them to machine translation proposals for new text. “While AI is good at quickly suggesting translation drafts, only a human with in-depth knowledge of the source and target languages can analyze lexical and semantic nu- ances and ensure that the meaning of the translation is identical,” says project leader Prof. Dr. Josef van Genabith, outlining the benefits. The project team at DFKI investigated how translation environments can support multimodal input and con- sider the cognitive aspects of post-editing. They also addressed the question of how automatic post-editing helps to avoid recurring errors. The researchers created a translation environment through a user-centered design process. An evaluation with pro- fessional users showed that these interaction modalities are good extensions to mouse & keyboard, with pen and touch input proving suitable for deletion and reordering tasks. At the same time, voice commands and multimod- al combinations of select & speak work well for substitu- tions and insertions. Robust approaches to automatically estimate the altered cognitive load during post-editing were designed to better understand whether and when MT tends to help or hinder the work process. They demonstrated that multimodal measures of eye-, heart-, and skin-based data can be used to adapt translation environments to cognitive load. MMPE, led by Prof. Dr. Antonio Krüger and Prof. Dr. Josef van Genabith, was funded by the German Research Foun- dation (DFG) for a period of three and a half years, ending on December 31, 2020. The MMPE project is now available as open-source on Github: https://github.com/NicoHerbig/MMPE More information https://mmpe.dfki.de https://github.com/NicoHerbig/MMPE https://youtu.be/H2YM2R8Wfd8 Contact Nico Herbig Research Department Cognitive Assistants email@example.com +49 681 85775 5368 Prof. Dr. Josef van Genabith Head of Research Department Multilinguality and Language Technology firstname.lastname@example.org +49 681 85775 5287
CEF AT TOOLS AND SERVICES TRANSLATION TOOLS FROM AND FOR EUROPE 37 1 2 0 2 / 1 S W E N I K F D The European Union firmly anchored the principle of multilingualism, reflecting its linguistic diversity, in its Charter of Fundamental Rights. Modern language techno- logies help to realize this promise in everyday activities. The aim of “CEF AT Tools and Services” (Connecting Europe Facility – Automated Translation) is to identify and provide multilingual tools and services in support of the ministries, public services, and SMEs in overcoming language barri- ers. This project was initiated by the EC in 2017, with a term ending in April 2021. As the lead manager, DFKI's Multilin- guality and Language Technology research unit partnered with CrossLang, ELDA, IDC, ILSP, and Tilde to develop the multilingual tools, concepts, and analyses required for a multilingual Europe. “CEF AT Tools and Services significantly raised the profile of European language technology vendors and demonstrated the broad spectrum and potential of language technolo- gy 'made in Europe.'” In the process, it became clear that continuing support of EU companies is essential. The re- search conducted in the project provides the basis for ac- tively shaping the future of the European translation service called eTranslation," said Dr. Andrea Lösch, project leader. DFKI and its partners organized conferences, language tech- nology workshops, and meetings with public service pro- fessionals, while also conducting several detailed surveys with stakeholders from all EU member states, plus Norway and Iceland. The purpose was to provide comprehensive information to European authorities about the possibilities and promise inherent in the use of language technologies and to create a forum for mutual exchange. A comprehensive requirements study was conducted to determine the needs of Europe's SMEs and public servic- es and to identify ways to expand the existing machine translation tool, eTranslation. The study revealed that users wanted a wider choice of languages (e.g., Arabic) and subject areas (e.g., business and finance, or technology and science). There was also interest expressed in the intro- duction of new features, like translation of text in images and automatic summarization. The first engines for non- EU languages (Russian, Chinese, and Turkish) were quickly released in 2020. According to a CEF marketing study for language technol- ogies, an annual growth rate of 10% is expected in the EU. However, the study also warned that with more than 500 European language technology firms, there is a danger of market fragmentation – a problem for the development of the European language technology market. The project also produced a “Catalog of CEF eTranslation Services.” It lists more than 670 tools and technologies from European service providers that support the work of public services and SMEs in Europe, including speech recognition and tools for anonymization and computer- aided translation. More information https://cef-at-service-catalogue.eu https://ec.europa.eu/cefdigital/wiki/display/ CEFDIGITAL/eTranslation Contact Dr. Andrea Lösch Research department Multilinguality and Language Technology email@example.com +49 681 85775 5285
38 D F K I N E W S 1 / 2 0 2 1 TRANSFER OF CONTROL AMONG AUTONOMOUS AGENTS TRACTAT PROJECT ENDS SUCCESSFULLY Advanced autonomous systems like robots, vehicles, and soft-bots can be employed in complex tasks in many different areas. Still, many activities are best performed by people, and there are situations that require human intervention. However, the same also applies the other way around. In the future, the temporary hand-off of control of a task to another agent will enable human and AI capabilities to mesh, and this will solve a wide variety of tasks in the work environment and in everyday life. The aim of project TRACTAT (Transfer of Control (ToC) between Autonomous Systems and Humans), sponsored by the Federal Ministry of Education and Research (BMBF), is to lay the foundation for a new frictionless and effective way of transferring control between autonomous sys- tems and humans within a cyber-physical environment. The big breakthrough came with the creation of a new basis for a generation of interactive platforms that fa- cilitate the control transfer between people and tech- nical agents. This requires real-time, multi-modal pres- entations of the current situation. The core activities in planning autonomous systems are machine learning and algorithmic action planning. The project ran for a term of three years, with a funding volume of 1.8 million euros, and was completed on September 30, 2020. When planning an autonomous system, the main chal- lenges are the design, implementation, and integration. Studies were performed to determine how aspects like the type, the point in time, and the content of the pres- entation would impact factors like ergonomics, safety, and user trust in the system. The resulting ToC frame- work, to include the parallel development of a world One of the TRACTAT system demonstrators designed and tested at the Human-Robot Collaboration 4.0 Innovation Lab. model, was developed along three selected application domains: INDUSTRIE 4.0, retail, and autonomous driving. The ToC model is validated using prototypes that map the individual steps in the basic control transfer process. The demonstrators developed for the three application domains show the interaction that takes place during a transfer of control from system to human and from hu- man to system. The project looked at user interface de- sign procedures that facilitate the proactive handover of control. The resulting dialog system supports various types of multi-modal references and other phenom- ena, for example, referencing external objects during the control transfer process. An exchange of informa- tion about the ToC and the world state takes place, for example, to select the camera that provides the optimal picture for the user. The implementation was performed at three of DFKI‘s Innovation Labs: MRK 4.0, Innovative Retail Laboratory (IRL), and Advanced Driver Assistance Systems (ADAS) and successfully applied many years of experience and expertise from related projects like Hybr-iT, MADMACS, and the BaSys family of projects. Work on the challeng- ing subject of control transfer will be continued in Pro- ject CAMELOT. More information https://tractat.dfki.de https://www.dfki.de/en/web/research/projects-and- publications/projects-overview/projekt/camelot/ Contact Dr. Michael Feld Research Department Cognitive Assistants firstname.lastname@example.org +49 681 85775 5328
40 N E W S I N B R I E F D F K I N E W S 1 / 2 0 2 1 hello2AI – Hybrid Conference on Artificial Intelligence This 3-day event is planned for May 28-30, 2021, online and, per- haps, with some in-person presence in Saarbrücken, Trier, Birken- feld, Kaiserslautern, and Berlin, depending on the infection rate. Hello2AI gives amateurs and professionals a forum to introduce creative ideas, to develop prototypes, and to have discussions with experienced mentors. An expert jury judges the best ide- as. The event is part of the GISplus project under the terms of the EXIST business startup grant program of the German Fed- eral Ministry for Economic Affairs and Energy (BMWi). DFKI, as a project partner, brings its expertise in the field of AI applications. Prof. Philipp Slusallek joins the Advisory Board at Z-Inspection as its newest member Prof. Roberto V. Zicari, Frank- furt Big Data Lab and Jo- hann Wolfgang Goethe-Uni- versity Frankfurt, started the initiative to contribute to the so-called “Mindful Use of AI” (#MUAI). The aim is to review AI systems for any potential ethical conflict and to evalu- ate the trustworthiness of AI. The Advi- sory Board is composed of more than 40 international experts and advises Z-Inspection in scientific and strate- gic matters. More information and event registration: www.hello2ai.de More information: http://z-inspection.org Prof. Slusallek appointed to the scientific council of the “Institute for Advanced Studies (IAS)” at the University of Luxembourg IAS Luxembourg was founded in 2020 for the purpose of promoting interdisciplinary research at the Universi- ty of Luxembourg, strengthening the university research community, improving the university‘s networks, and attracting young, talented research scholars. IAS is supported and advised by the 13 members of the Scientific Council. As a Council Member, Prof. Slusallek participates in the annual evaluation of interdisciplinary projects and the work of the doctoral candidates. Oth- er duties include commenting on ongoing key activities at IAS and contributing to the international visibility and reputation of IAS Luxembourg. Best Paper Award at the “LatinX in AI Research Workshop @ NeurIPS 2020” Dr. Matias Valdenegro Toro (DFKI) and Luis Octavio Arriaga-Camargo (University of Bremen) received the award for Best Paper for their presentation of “Unsu- pervised Difficulty Estimation with Action Scores” as part of the NeurIPS 2020 Conference (International Conference on Neural Information Processing Sys- tems), hosted virtually in December 2020. Luis Octavio Arriaga-Camargo, University of Bremen. The paper presents a simple method for computing an “action score” that is useful in estimating the dif- ficulty of samples in data sets. The method helps to create better and fairer models and is useful for find- ing biases and problems in machine learning models and datasets. “LatinX in AI” (LXAI) is an organization that supports AI research in Latin America and organizes workshops in the context of international conferences. Dr. Matias Valdenegro Toro, DFKI. News in brief
41 F E I R B N I S W E N 1 2 0 2 / 1 S W E N I K F D Robotics Innovation Center (RIC) presents EU project funding program for robotics in industry and maintenance Robots can prove their potential value where humans are at risk, precision is required, and fast action is needed – for example, for inspection and main- tenance tasks. As a component of the EU-sponsored RIMA project, DFKI cooperates with small and medium-sized companies that are interested in robotics to maintain and service their plants. In addition to project funding, DFKI‘s RIC also offers training, administrative services, and the use of its test facilities. The RIC manages one of the 13 Digital Innovation Hubs (DIH) funded by the EU over four years, which promotes the exchange of ideas be- tween research and industry. Prof. Robert Wille receives coveted ERC Consolidator Grant Best Paper Awards to DFKI‘s Cyber- Physical Systems research department For his work in the field of quantum computing, Prof. Dr. Robert Wille received the prestigious Consolida- tor Grant awarded by the European Research Council. Prof. Wille, from the Johannes Kepler University Linz and the Software Competence Center Hagenberg, is also associat- ed with the DFKI Cyber-Phys- ical Systems department. His expertise is in develop- ing methods for improving the performance of quan- tum computers. The research project explores methods for the simulation and proof of correctness of corresponding quantum programs and is funded by the ERC with two million euros. The Best Paper Award at “Euromicro Conference on Digital System Design” (DSD), held in August 2020, was awarded to only one research team. The team members were Prof. Dr. Rolf Drechsler and Dr. Philipp Niemann, both from the University of Bremen and DFKI‘s Cyber-Physical Systems department, Alexandre Almeida from Sao Paulo State Uni- versity in Brazil, and Prof. Gerhard Dueck from the Univer- sity of New Brunswick in Canada. The prize was awarded for their publication “Design Space Exploration in the Map- ping of Reversible Circuits to IBM Quantum Computers.” In September 2020, another team led by Prof. Drechsler was honored with Best Paper Award at the “Forum on Specifi- cation & Design Languages (FDL).” Scientists Dr. Vladimir Herdt (DFKI), Prof. Dr. Daniel Große (Johannes Kepler Univer- sity Linz and DFKI), Eyck Jentzsch (MINRES® Technologies), and Rolf Drechsler received the prize for their publication “Efficient Cross-Level Testing for Processor Verification: A RISC-V Case-Study,” in which they presented a new test approach for processor verification. Book tips for young fans of robotics “Alles über Roboter” (Everything about Robots) is the recently published title of Volume 47 in the Ravensburger children‘s books series “Why? For what? What for?” Author Andrea Erne and illustrator Markus Humbach chose to feature, among others, the systems at the DFKI Robotics In- novation Center. In the book, readers will find the six-legged MANTIS, the space robot SherpaTT, the ape-like Charlie and the Coyote III Rover. “Finja forscht – Der rätselhafte Roboter” (Finja investigates – The curi- ous robot) by author Isabell Harder and illustrator Lea Fröhlich, is a fic- tional video chat with Prof. Dr. Rolf Drechsler, Chair of the Computer Architecture Working Group at the University of Bremen and Head of DFKI‘s Cyber-Physical Systems research, in which he answers questions about “Artificial Intelligence” posed by Finja and Malik, the two charac- ters in the children‘s book.
42 D F K I P R O F I L E D F K I N E W S 1 / 2 0 2 1 German Research Center for Artificial Intelligence Company Profile The German Research Center for Artificial Intelligence (DFKI) was found- ed in 1988 as a non-profit public-private partnership. It has research facilities in Kaiserslautern, Saarbrücken and Bremen, a project office in Berlin, a Laboratory in Niedersachsen and branch offices in Lübeck, St. Wendel and Trier. In the field of innovative commercial software technology using Artificial Intelligence, DFKI is the leading research center in Germany. Based on application oriented basic research, DFKI develops product functions, prototypes and patentable solutions in the field of infor- mation and communication technology. Research and development projects are conducted in 24 research departments, nine competence centers and eight living labs.Funding is received from government agencies like the European Union, the Federal Ministry of Education and Research (BMBF), the Federal Ministry for Economic Affairs and Energy (BMWi), the German Federal States and the German Research Founda- tion (DFG), as well as from cooperation with industrial partners.Twice a year, a committee of internationally renowned experts (Scientific Ad- visory Board) audits the progress and results of state-funded projects. Apart from the state governments of Rhineland-Palatinate, Saarland and Bremen, numerous renowned German and international high- tech companies from a wide range of industrial sectors are repre- sented on the DFKI supervisory board. The DFKI model of a non-profit public-private partnership (ppp) is nationally and internationally consid- ered a blueprint for corporate structure in the field of top-level research. DFKI is actively involved in numerous organizations representing and continuously advancing Germany as an excellent location for cut- ting-edge research and technology. Far beyond the country’s borders DFKI enjoys an excellent reputation for its academic training of young scientists. At present, approx. 670 highly qualified researchers, admin- istrators and 450 graduate students from more than 65 countries are contributing to more than 250 DFKI research projects. DFKI serves as a stepping stone to leading positions in industry and successful ca- reers as founders of spin-off companies. Over the years, more than 140 staff members have been appointed professors at universities in Germany and abroad. More information www.dfki.de Contact Reinhard Karger, M.A. Corporate Spokesperson German Research Center for Artificial Intelligence GmbH Saarland Informatics Campus D3 2, 66123 Saarbrücken, Germany email@example.com +49 681 85775 5253 Established 1988, non-profit organization (public-private partnership) Executive Board Prof. Dr. Antonio Krüger Supervisory Board Dr.-Ing. Gabriël Clemens, VSE AG (Chair) Dr. Susanne Reichrath, Representative of Saarland’s Minister President for Higher Education, Science and Technology (Vice Chair) Shareholders Accenture, Airbus Group, Bilfinger Digital Next GmbH, BMW AG, Cerence GmbH, CLAAS KGaA mbH, Daimler AG, Deutsche Börse AG, Deutsche Messe AG, Deutsche Telekom AG, Empolis Informati- on Management GmbH, Fraunho- fer Gesellschaft e.V., Google Inc., HARTING AG & Co. KG, Intel Corporation, John Deere GmbH & Co. KG, KIBG GmbH, Microsoft Deutschland GmbH, Münchener Rückversicherungs-Gesellschaft Aktiengesellschaft in München, NVIDIA GmbH, RICOH Company Ltd., Robert Bosch GmbH, ROSEN Swiss AG, SAP SE, Sartorius Ventures GmbH, Schwarz-Gruppe, Software AG, Technische Univer- sität Kaiserslautern, Universität Bremen, Universität des Saarlan- des, Volkswagen AG, VSE AG, ZF Friedrichshafen AG International Scientiﬁc Advisory Board Bi-annual evaluation of publically funded projects: Prof. Dr. Andreas Butz, Ludwig-Maximilians-Universität, München (Chairman) Key Figures 2020 Annual Budget: ca. € 64,6 million Total Assets: ca. € 171,3 million Professional staff: 670 Graduate student staff: 450 Profile
HUMAN-CENTRIC AI INDUSTRIE 4.0, Digital Twins, Semantic Product Memories • Smart Data – Intelligent Analytics for Massive Data • Wearable Computing and Interactive Textiles • Deep Learning and Machine Learning • Knowledge Management and Document Analysis • Softbots, Digital Assistants, and Chatbots • Educational Technologies • Verification and Evaluation of Safety-critical Applications • Cognitive Social Simulation • Information Extraction and Intelligent Web Retrieval • Multiagent Systems • Experience-based Learning Systems • Visual Computing and Augmented Vision • Mobile and Collaborative Robotic Systems • Multimodal User Interfaces and Autonomous Systems • Shopping Assistance and Intelligent Logistics • Safe and Secure Cognitive Systems and Intelligent Security Solutions • Ambient Intelligence and Assisted Living • Driver Assistance Systems and Autonomous Driving • Cyber-physical Systems • Multilingual Technologies and Language Understan- ding • Business Process Management and Smart Services • Affective Computing • AI in Medicine and Healthcare Kaiserslautern Site Trippstadter Straße 122 D-67663 Kaiserslautern +49 631 20575 0 Saarbrücken Site Saarland Informatics Campus D3 2 D-66123 Saarbrücken +49 681 85775 0 Bremen Site Robert-Hooke-Straße 1 D-28359 Bremen +49 421 17845 0 Laboratory Niedersachsen Berghoffstraße 11 D-49090 Osnabrück +49 541 386050 0 www.dfki.de firstname.lastname@example.org