Professor Wolfgang Wahlster is a pioneer of AI in Germany and Europe as a founding director of the DFKI. He has served as an elected President of three international AI organizations: IJCAII, EurAI, and ACL. He is an elected Fellow of AAAI, EurAI, and GI. He laid some of the foundations for multimodal dialog systems, user modelling, and speech-to-speech translation cyber-physical production systems for the fourth industrial revolution (Industrie 4.0), a concept that he coined in 2010. Wahlster is a member of the Nobel Prize Academy in Stockholm, the German National Academy Leopoldina and three other prestigious academies. For his research, he has been awarded the German Future Prize, and the Grand Cross of Merit by the Federal President of Germany. (for more info see: https://www.wolfgang-wahlster.de/)

In the next decade of Industry 4.0 a new generation of AI technologies will take smart factories to a new level. Large Language Models (LLMs) will be complemented by Large Process Models (LPMs) and Large Action Models (LAMs), so that generative AI models not only predict what to say or visualize next, but also what to do next with explanations of why these actions make sense.
Although deep learning is the most powerful machine learning method developed to date, it has already reached its inherent limits in many industrial application domains. It must be combined with various symbolic approaches in new system architectures. This leads to hybrid LxM (x=L,P, or M) technologies that use holonic multiagent architectures for combining neural approaches with symbolic reasoning technologies such as constraint solving, physics-based simulation and terminological reasoning in knowledge graphs.

Dr.-Ing. Daniel Porta received a diploma in Computer Science from Saarland University in 2007 and his doctoral degree in 2017. Joining DFKI’s Cognitive Assistants research department already as a student in 2004, he is now a Senior Researcher leading a research group on industrial AI.

Digital twins collect information on an asset over its entire life cycle and provide it in a standardised way for a wide range of applications. The course will introduce to digital twin architectures based on Asset Administration Shells and further sound abstraction layers for future-proof Industrie 4.0 infrastructures. It then discusses several industrial use cases in terms of AI-based applications at different life cycle phases.

Kevin Baum, a computer scientist (M.Sc.) with a doctorate in philosophy, is the head of the Center for European Research in Trusted Artificial Intelligence (CERTAIN) and deputy head of the Neuro-Mechanistic Modeling (NMM) research department at the German Research Center for Artificial Intelligence (DFKI). In various interdisciplinary research projects, he has researched primarily on the sense and nonsense of transparency and explainability requirements for AI systems with regard to societal desiderata such as recognizing unfairness, the effectiveness of human oversight and enabling moral responsibility. He developed the award-winning lecture Ethics for Nerds, is and has been a member of various ethics committees, and is part of Algoright e.V., the interdisciplinary non-profit think tank for good digitalization and science communication.

“Progress in the field of AI is breathtaking. Large models, foundation models, multimodality: all this not only opens up a wide range of new possibilities, be it in code generation with LLMs or in robotics via Large Action Models, but also raises new societal and ethical challenges. In his provisional keynote, Kevin Baum provides an overview of current normative challenges, sorts out loose threads, and outlines some resulting research opportunities.”

Pia Bideau is a researcher at the THOTH team at Inria Grenoble and holds a junior research chair position for “Perception and Interaction” at MIAI Grenoble Alpes. Before joining Inria in October 2023, Pia Bideau was postdoctoral researcher at the robotics lab at Technical University Berlin and the Excellence Cluster “Science of Intelligence” (SCIoI). She received her PhD from the University of Massachusetts, Amherst, advised by Prof. Erik Learned-Miller. Her thesis proposed novel approaches towards segmenting independently moving objects from noisy optical flow fields.

Pia Bideau has co-organized the workshop “What is motion for?” at ECCV 2022 has engaged in numerous projects focused on research-oriented teaching. In the past she received a best paper award at the ECCV 2018 Workshop: What is optical flow for? for her paper “MoA-Net: Self-Supervised Motion Segmentation” and scholarships from DAAD for academic education abroad and from BMBF for excellent academic achievements during her Master studies.

Distance estimation is an essential part of scene recognition and orientation, allowing agents to move in a natural environment. In particular, when humans or animals move in teams, they seem to be capable of doing this – moving together as a whole without colliding or bumping into each other. Different sensor systems but also different strategies of movement enable agents to localize themselves relative to their neighbors or neighboring objects.

This course provides an introduction into analytical and learning based approaches for distance estimation. While there are several cues to extract information about distance the focus of this course lies on object appearance and its relative size. Objects appearing at greater distance will appear smaller than objects nearby. This is one of the fundamental principles of perspective projection. A classical object detector (YOLOv5 small/nano) will be extended with the ability to estimate distance. When does a system benefit from learning? When should estimates be computed following known physical principles instead of being learned from data? This part of the assignment focuses on implementing both solutions to distance estimation – the analytical computation and a multilayer perceptron (MLP). It involves analyzing the advantages and disadvantages of each approach and ultimately deciding which algorithm to deploy on a real robot. If time permits, we will delve into ongoing research addressing the challenges of distance estimation for behavior analysis, specifically focusing on reconstructing speed and 3D trajectories of animals and humans.

Prerequisites: basic python programming skills.

Gianluca Rizzello was born in Taranto, Italy, in 1987. He received the master’s (Hons.) degree in control engineering from Polytechnic University of Bari, Bari, Italy, in 2012. He received his Ph.D. in Information and Communication Technologies from Scuola Interpolitecnica di Dottorato, a joint program between Polytechnic Universities of Torino, Bari, and Milano, Italy, in 2016. After his doctoral studies, he joined Saarland University, Saarbrücken, Germany, first with the role of a postdoc researcher and Group Leader Smart Material Modeling and Control (2016-2019), and subsequently as Assistant Professor in Adaptive Polymer Systems (2020 – present). His research interests include development, modeling, and control of soft robotic and mechatronic systems based on unconventional drive technologies, such as smart materials.

While traditional robots are essentials in many industrial tasks, they show limitations when performing certain tasks involving safe interaction with humans or exploration of complex unstructured environments. Taking inspiration from animals and other biological systems, the field of soft robotics offers a possible means to develop intelligent machines that can interact with their environment in ways rigid robots cannot. Soft robots benefit from the presence of elastic and soft elements that enhance their adaptability and versatility with respect to their environment, thus allowing to close the gap between traditional rigid robots and biological systems. Integrating soft features into robotic systems, however, involves several challenges in terms of system design, component selection, modeling, and control. This lecture aims at providing a general introduction to the field of soft robotics. The lecture will cover both hardware and software aspects of soft robots, ranging from soft design principles, soft actuators, and soft sensors, to challenges posed by modeling and control of soft robots. For each one of those areas, the main results from the state of the art, major challenges, and research opportunities will be illustrated. The presentation of the topics will be accompanied by several examples from the soft robotic literature.

Marie-Odile Berger is INRIA Research Director at the “Centre INRIA de l’Université de Lorraine” and currently heads the TANGRAM Computer Vision Group. Her research interests include computer vision, artificial intelligence with an application focus on augmented reality tasks requiring high-precision localization, both in classical environments and in medical imaging. Her research has led to several theoretical and practical results in the areas of matching and 3D tracking, reconstruction and visual perception. She has published more than 140 papers in conferences and journals. http://members.loria.fr/moberger

Like many other fields, image-based localization methods have benefited greatly from the emergence of convolutional networks (CNN). In this course, after describing the main principles of methods using AI for localization, we will focus on methods based on high-level features derived from CNN. In particular, we’ll look at methods that use objects detected in images as landmarks for localization. Such methods, based on the use of a generic object detector, have many advantages: they avoid systematic re-training of algorithms for new scenes, do not require a precise model of the scene and have very good accuracy.

Xavier Hinaut is Research Scientist in Bio-inspired Machine Learning and Computational Neuroscience at Inria, Bordeaux, France since 2016. He received a MSc and Engineering degree form Compiègne Technology University (UTC), FR in 2008, a MSc in Cognitive Science & AI from EPHE, FR in 2019, then his PhD of Lyon University, FR in 2013. He is a member (Vice Chair) of IEEE CIS Task Force on Reservoir Computing. His work is at the frontier of neurosciences, machine learning, robotics and linguistics: from the modeling of human sentence processing to the analysis of birdsongs and their neural correlates. He both uses reservoirs for machine learning (e.g. birdsong classification) and models (e.g. sensorimotor models of how birds learn to sing). He manages the “DeepPool” ANR project on human sentence modeling with Deep Reservoirs architectures and the Inria Exploratory Action “BrainGPT” on Reservoir Transformers. He leads ReservoirPy development: the most up-to-date Python library for Reservoir Computing. https://github.com/reservoirpy/reservoirpy He is also involved in public outreach, notably by organising hackathons from which fun projects with reservoirs came out (ReMi project on reservoir generating MIDI and sounds).

Language involves several levels of abstraction, from small sound units like phonemes to contextual sentence-level understanding. Large Language Models (LLMs) have shown an impressive ability to predict human brain recordings. For instance, while a subject is listening to a book chapter from Harry Potter, LLMs can predict parts of brain imaging activity (recorded by functional Magnetic Resonance Imaging or Electroencephalography) at the phoneme or word level. These striking results are likely due to their hierarchical architectures and massive training data. Despite these feats, they differ significantly from how our brains work and provide little insight into the brain’s language processing. We will see how simple Recurrent Neural Networks like Reservoir Computing can model language acquisition from limited and ambiguous contextual data better than LSTMs. From these results, in the BrainGPT project, we explore various architectures inspired by both reservoirs and LLMs, combining random projections and attention mechanisms to build models that can be trained faster with less data and greater biological insight.

Xiaomei Xu joined ZeMA – Zentrum für Mechatronik und Automatisierungstechnik gemeinnützige GmbH as a research assistant in the Robotics group in spring 2020 after graduating from RWTH Aachen University. She focuses on developing mathematical algorithms for 3D cameras and robotics applications. Xiaomei’s work has been presented at conferences such as the CIRP Web Conference 2020, IEEE CASE 2021&2024, IEEE ICSC 2022, CIRP CATS 2024, and CIRP ICME 2024.

In the 90-minute “Hands-on Robotik-Kurs”, we divide the group activity into two segments. The first segment, lasting 30 minutes, provides a Quick Guide for Siemens Tecnomatix (Software User Interface and basic function). This guide explains how to build a virtual environment for robotic simulation applications in welding, deburring, and painting. The second segment, lasting 60 minutes, involves robot path planning based on the manufacturing geometry for welding, polishing and painting processes.

Dr.-Ing. Tim Schwartz studied computer science and computational linguistics at Saarland University. He received his PhD with the thesis “The Always Best Positioned Paradigm for Mobile Indoor Applications” in 2012. Since 2016, he is the head of the human-robot communications group and leads the German-Czech Innovation Lab for Human-Robot Collaboration MRK 4.0.

In this tour, we will show you around in our german-Czech Innovation Lab for Human-Robot Collaboration in Industrie 4.0, or MRK 4.0 Lab for short. As the name implies, we focus on Human-Robot Collaboration. In extension we also deal with human-robot communication, the orchestration of hybrid teams (i.e. teams consisting of humans, robots and software agents) and practical applications of Industrial AI, Asset Administration Shells, Digital Twins and general Industrie 4.0 topics. Human-Robot communication is not necessarily limited to spoken or written language, but includes all sorts of modalities: from more traditional control units, over manual teach-in to Augmented and Virtual Reality etc. We will show you practical examples from different projects, we are currently working on or have been working on in the past, encouraging questions and discussions throughout the whole experience. 

Martin Suda is a senior researcher at the CIIRC institute of the Czech Technical University in Prague and the head of the Automated Reasoning Group there. He is also a part-time research scientist at Filuta.ai. His primary research interest is automated theorem proving and how it can be boosted through the techniques of machine earning. He is one of the main developers of the award-winning automatic theorem prover Vampire.

We will provide an overview of the state-of-the-art technology in logic-based reasoning, ranging from propositional satisfiability and satisfibility modulo theories to automatic and interactive theorem proving. The corresponding tools, often referred to as “solvers”, find many applications in areas such as hardware verification (chip design), software verification (program correctness) or automation of math. We will also discuss how machine learning and, in particular, neural networks enter the picture of the development of such solvers, able to automatically help discover new guidance heurstics, so necessary for fighting the inherent combinatorial explosion. Finally, we will also contemplate the opposite direction for field synergy: Couldn’t logic-based tools help us eliminate errors from neural networks’ outputs, most notably guard as against LLM hallucinations?