Scientists develop new machine learning method that can make robots safer - New method provides simpler and more intuitive models of physical situations
Text: Dr. Elisabeth Guggenberger, IST. Understanding how a robot will react under different conditions is essential to guaranteeing its safe operation. But how do you know what will break a robot without actually damaging it? A new method developed by scientists at the Institute of Science and Technology Austria (IST Austria) and the Max Planck Institute for Intelligent Systems (MPI for Intelligent Systems) is the first machine learning method that can use observations made under safe conditions to make accurate predictions for all possible conditions governed by the same physical dynamics. Especially designed for real-life situations, their method provides simple, interpretable descriptions of the underlying physics. The researchers will present their findings tomorrow at this year’s prestigious International Conference for Machine Learning (ICML).
Modeling anguilliform swimming, robotic swimming
Anguilliform swimming is a locomotor mode that has been preserved for millions of years and is used by many different animals such as eels, lampreys, salamanders or leeches. Aside from highly energy efficient locomotion characteristics, studying the nervous system of anguilliform swimmers has shed light on many aspects of neural control in the past, e.g. on the neural structure of C. elegans. In addition, these studies have helped to reveal mechanisms in the spinal cord (nerve cord in invertebrates) responsible for locomotion. In this context researchers have for instance discovered so-called Central Pattern Generators (CPGs), which represent distributed neural networks that spontaneously produce rhythmic activity. Besides these open-loop pattern generators the role of corresponding local feedback mechanisms has been studied much less. While it is hard to analyze and understand such feedback loops in living animals, due to the highly interconnected nature of the sensorimotor apparatus, comprehensive models offer a great way to identify key mechanisms and their contribution in locomotion control. In this talk, I will present our recent modelling work on local feedback based anguilliform swimming. By using distributed force feedback and phase oscillators in a simulated viscoelastic body and a hydrodynamic environment, we were able to discover new travelling wave generation mechanisms in elongated swimmers. These results highlight the importance of neural feedback mechanisms and open a new venue for decentralized control of anguilliform swimming robots.
Das Bundesministerium für Bildung und Forschung unterstützt Projekt des Max-Planck-Instituts für Intelligente Systeme innerhalb des Deutsch-Schwedischen Großforschungsprojekts Röntgen-Ångström-Cluster mit 1,2 Millionen Euro. Ein Drittel davon gehen für das Teil-Projekt DynaMAX nach Stuttgart.
Composite surface has features that can move microparticles, mix droplets, repel biofilms and more.
Want to contribute to world-leading research in areas such as machine learning, computer vision, robot dynamics and control, micro- and nano-robotics, or haptics? Then sign up for the International Max Planck Research School for Intelligent Systems or IMPRS-IS now!
A prestigious junior scientist fellowship is awarded to the MPI-IS researcher
Several scientists from the Max Planck Institute for Intelligent Systems have recently given keynote speeches at some of the most important conferences in their fields.
Dr. Hamed Shahsavan receives a prestigious Postdoctoral Fellowship from the Natural Sciences and Engineering Research Council (NSERC) of Canada, the country´s federal funding agency for university-based research and student training in natural sciences and engineering. The smart materials engineer choses the Max Planck Institute for Intelligent Systems for his research stay because of the state-of-the-art facilities provided there.
Beobachtung magnetischer Tröpfchen große Bedeutung für die magnetische Datenprozessierung
Forscher am Max-Planck-Institut für Intelligente Systeme in Stuttgart konnten mit Hilfe eines Röntgenmikroskops bei der Bildung von magnetische Tröpfchen ein völlig unerwartetes Verhalten beobachten. Wenn der Strom über einen Nanokontakts durch die magnetische Schicht fließt, breitet er sich wesentlich weiter aus, als die Ausdehnung des Nanokontakt es zulassen sollte. Bisher waren Forscher davon ausgegangen, dass nur die Fläche unterhalb des Nanokontakts reagiert. Doch Experimente haben die Wissenschaftler eines Besseren belehrt. Ein spannendes Phänomen in der Grundlagenforschung und von großer Bedeutung für die magnetische Datenprozessierung.