Linking Neuroscience to Robotics: How Complex Machines Benefit from Brain Theory

Transcription

1 German-Japanese Science and Innovation Forum 2010 Wednesday, October 6 th, 2010, Roppongi academyhills 49, Tokyo Workshop 5 Venue: Collaboration Room 1+2, 14:00 17:00 Linking Neuroscience to Robotics: How Complex Machines Benefit from Brain Theory Organised by: German Research Foundation DFG Animals can perceive their world and perform complex motor actions in response to the task at hand; humans in addition are able to comprehend and modify their world in a cognitive way. All these achievements are entirely based on the parallel processing of billions of neuron in the brain. Through theoretical neuroscience an increasingly better understanding of brain function has emerged during the last decades. Algorithms have been developed to simulate perception and action and also to some degree cognitive properties. While fundamentally based on the biophysics of the brain s wetware, it has become possible to derive abstractions from these models by which complex functions are emulated in a way compatible to the hardware of robots and other technical systems. Currently this leads to improved sensori-motor control and increased adaptivity (learning) at machines hence to novel avenues to be increasingly exploited and of possible high potential also for industrial applications. The goal of this workshop is to discuss these aspects between experts from neuroscience and robotics. 1

2 Programme Welcome Dr. Iris Wieczorek, Director DFG Office Japan Chair Professor Dr. F. Wörgötter, Georg-August-University Göttingen Anthropomorphic Biological Equipment (ABE) in the Brain and Anthropomorphic Artificial Equipment (AAE) of Robots Professor Dr. Yoshihiko Nakamura, The University of Tokyo Can Cognitive Developmental Robotics Make a Paradigm Shift? Professor Dr. Minoru Asada, Osaka University Humanoid Robotics and Neuroscience Dr. Tamim Asfour, Institute for Anthropomatics, Karlsruhe Institute of Technology (KIT) From Self-organised Criticality in Neural Systems to Self-organisation of Behaviour in Robots Professor Dr. J. Michael Herrmann, Institute for Perception Action and Behaviour, School of Informatics, University of Edinburgh, U.K. Extraction of Common Representation Shared by Neural Activities and Movements, and Its Application to Brain-Robot Interface Professor Dr. Jun Morimoto, ATR Labs, Nara Grasping and the Brain - A Hands On Approach from Robotics Professor Dr. Helge Ritter, Cognitive Interaction Technology Excellence Cluster (CITEC), Bielefeld University Robots under Neural Control Professor Dr. F. Wörgötter, Georg-August-University Göttingen The scientific presentations will then be followed by an Open Discussion Forum with all speakers. 2

3 The German Research Foundation (Deutsche Forschungsgemeinschaft, DFG) The DFG is the self-governing organisation for science and research in Germany. It serves all branches of science and the humanities. The DFG is an association under private law. Its membership consists of German research universities, non-university research institutions, scientific associations and the Academies of Science and the Humanities. The DFG receives the large majority of its funds from the states and the Federal Government, which are represented in all Grants Committees. At the same time, the voting system and procedural regulations guarantee science-driven decisions. The chief task of the DFG is to select the best research projects by scientists and academics at universities and research institutions on a competitive basis and to finance these projects. Projects are presented by scientists and academics or by universities in a proposal, which is then evaluated by voluntary reviewers exclusively according to scientific criteria. DFG funding guarantees quality-based differentiation in the German research system. The DFG fosters international collaboration, the mobility of researchers, and the internationalisation of German universities. With its liaison offices in Washington, New York, Moscow, New Delhi and Tokyo, as well as the Sino-German Center for Research Promotion in Beijing, the DFG has expanded its presence in other research regions around the world in recent years. It thus strengthens collaboration between researchers and their relationships to national funding organisations. The DFG office Japan opened in Tokyo in April 2009 in order to consolidate and intensify the multifaceted German-Japanese scientific collaborations. 3

4 Abstracts of Presentations Anthropomorphic Biological Equipment (ABE) in the Brain and Anthropomorphic Artificial Equipment (AAE) of Robots Professor Yoshihiko Nakamura, University of Tokyo Evolution of the human brain emerged by the pressure for communication skill and, for which reason the communication skill is specially targeted to anthropomorphism. One may call the total skill of the brain to perceive, recognize, understand and respond to the behaviors of an anthropomorphic system, the Anthropomorphic Biological Equipment. The natural and general human-machine interface can be established by developing the similar function on the machines side, which one may call the Anthropomorphic Artificial Equipment. Our research started from the mathematical model of mirror neurons and continues to acquisition of semiology of human behaviors based on technologies such as unsupervised segmentation, iterative clustering, and construction of state transition network. The current research interests target to connect the behavioral semiology to the semiology of a natural language to develop a statistical system that evokes mutual association. This talk introduces the scope of our research and overviews the direction of research on the Brain and Robots. Can Cognitive Developmental Robotics Make A Paradigm Shift? Professor Minoru Asada, Osaka University This talk discusses how cognitive developmental robotics (hereafter, CDR) can make a paradigm shift in science and technology. A synthetic approach is revisited as a candidate for the paradigm shift, and CDR is reviewed from this viewpoint. Focusing on the developmental process of self/others concepts, sociality developmental is modeled by CDR, and the relationship to the mirror neuron system is argued. Finally, a synthetic approach is proposed to emerge the concepts of self and others. A key idea is synchronization and desynchronization with objects, caregivers, and other agents. A trans-disciplinary approach appears to be a necessary condition and how to represent and design subjectivity' is an essential issue. 4

5 Humanoid Robotics and Neuroscience Dr. Tamim Asfour, Institute for Anthropometrics, Karlsruhe Institute of Technology (KIT) Building humanoid robots able to act and interact with the real world is a challenging task which requires coordinated and integrated research efforts that span a wide range of disciplines such as learning theory, control theory, human-machine-interaction, mechatronics, perception, biomechanics and neuroscience. These fields have usually been explored independently, leading to significant results within each discipline. This talk will present recent work on humanoid robotics focusing on grasping and manipulation in human-centered environments as well as on goal-directed imitation of human actions and discuss how a bidirectional interaction between robotics and neuroscience can contribute to the advancement of the field. Co-authors: Rüdiger Dillmann, Kai Welke, Martin Do, Pedram Azad, Nikolaus Vahrenkamp From Self-organised Criticality in Neural Systems to Self-organisation of Behaviour in Robots Professor Dr. J. Michael Herrmann, Institute for Perception Action and Behaviour, School of Informatics, University of Edinburgh In neural tissue avalanche-like activity waves of high diversity have been observed suggesting the interpretation of a self-exploration of its own dynamical options by the neural network. Neurally inspired control can benefit from this type of neural dynamics which may also relate to the redundant and multifarious structures underlying biological motor control. Starting from a model of self-organised criticality in neural system, we derive an algorithm that generates exploratory behaviour in robots enabling the robot to extract behavioural options for later usage in higher-level learning. In addition we discuss methods for guidance of the process of self-organisation of behaviour such as to bias the behaviour towards promising regions in the space of behavioural modes, to include background knowledge or to achieve more coherent representations. We show by example that algorithms for learning by self-organisation are particularly useful for robotic devices interacting with biological systems such as in prosthetics. Coworkers: Anna Levina, Georg Martius, Frank Hesse, Katja Fiedler, Theo Geisel 5

6 Extraction of Common Representation Shared by Neural Activities and Movements, and Its Application to Brain-Robot Interface Professor Jun Morimoto, ATR Labs, Nara In this talk, we introduce our approach to understand sensory-motor mechanisms in our brain. Recent advance of the multi-unit recording system allows us to simultaneously record monkey s neural activities from many neurons with capturing monkey s movements. In our approach, we try to extract low-dimensional features shared by the recorded neural activities and the captured movements using canonical correlation analysis (CCA). We discuss how the extracted low-dimensional common representation can be useful to construct a brainrobot interface (BRI), i.e., the interface to connect the brain to a robotic device. Grasping and the Brain A Hands On Approach from Robotics Professor Dr. Helge Ritter, Cognitive Interaction Technology Excellence Cluster (CITEC), Bielefeld University Today's robot hands are approaching some of the physical abilities of biological hands. The attempt to control such devices to partly replicate the abilities of biological hands can help to clarify computational tasks that are solved by brain circuits. Conversely, ideas about computational strategies of the nervous system for enabling manual operations can be put to novel tests and can inspire algorithmic approaches to robot hand control. This offers new opportunities for a tighter interaction between robotics and brain research. We discuss some issues arising from the emerging research field of manual intelligence", illustrate examples of brain-inspired" grasping strategies as well as learning approaches to refine algorithmic grasping strategies with human movement expertise, and point out some questions for future research. Robots under Neural Control Professor Dr. F. Wörgötter, Georg-August-University Göttingen Conventional robot control relies on methods from electrical engineering and control theory. This often leads to unwanted properties, because these controllers are stiff, inflexible, and cannot easily adapt ( learn ) to new situations. Here we will show several examples of robots, which operate under neuronal control, where signals from sensors are being processed by a 6

7 network to create motor-control signals directly from neuronal elements. The machines thereby develop fast responses and flexible behavior and can adapt by ways of simulated neuronal plasticity to new situations. Specifically we will demonstrate, how a small CPG, which operates initially in a chaotic domain can via neural chaos control be driven into repetitive patterns, which are being used to determine the walking gait of an 18DOF hexapod robot (Steingrube et al. Nature Physics, 2010). 7

8 Notes on Speakers Yoshihiko Nakamura, The University of Tokyo Yoshihiko Nakamura is a professor at Department of Mechano-Informatics, University of Tokyo. His research stems from the foundation of kinematics, dynamics, and control. Humanoid robotics, cognitive robotics, neuromusculoskeletal human model, biomedical systems, and their computational algorithms are his current fields of research. The publications of his group have received various academic awards including the King-Sun Fu Memorial Best Transactions Paper Award, IEEE Transaction of Robotics and Automation in 2001 and He received Most Active Distinguished Lecture Award in He currently serves as the Vice President of International Federation for the Promotion of Mechanism and Machine Science (IFToMM), and the Chairman of the Japan Council of IFToMM. Dr. Nakamura is a Foreign Member of Academy of Engineering Science of Serbia, and a Fellow of the Japan Society of Mechanical Engineers, the Robotics Society of Japan, and the World Academy of Art and Science. The University of Tokyo was established in 1877 as the first national university in Japan. As a leading research university, the University of Tokyo offers courses in essentially all academic disciplines at both undergraduate and graduate levels and conducts research across the full spectrum of academic activity. The university aims to provide its students with a rich and varied academic environment that ensures opportunities for both intellectual development and the acquisition of professional knowledge and skills. The University of Tokyo has a faculty of over 4,000 and a total enrollment of about 29,000, evenly divided between undergraduate and graduate students. As of 2006 there were 2,269 international students, and over 2,700 foreign researchers come annually to the university for both short and extended visits. The University of Tokyo is known for the excellence of its faculty and students and ever since its foundation many of its graduates have gone on to become leaders in government, business, and the academic world. Minoru Asada, Osaka University Minoru Asada received his B.E. (1977), M.E. (1979), and Ph.D. (1982) in Control Engineering from Osaka University, Osaka, Japan. Since April 1997, he has been a Professor at the Department of Adaptive Machine Systems, Graduate School of Engineering, Osaka University (Suita, Japan). Also, during August

9 and October 1987, he was a visiting researcher at the Center for Automation Research, University of Maryland (College Park, Maryland). He served as President of the RoboCup Federation ( ). In 2005, he was elected and continues to serve as research director of ASADA Synergistic Intelligence Project of ERATO (Exploratory Research for Advanced Technology, ) supported by the Japan Science and Technology Agency. Osaka University Under the motto "Live Locally, Grow Globally," Osaka University grapples with the challenges of education and research. 11 Schools with 10 Corresponding Graduate Schools: Letters, Human Sciences, Foreign Studies (School only), Law and Politics, Economics, Science, Medicine, Dentistry, Pharmaceutical Sciences, Engineering, Engineering Science Graduate School of Engineering The Graduate School of Engineering aims to develop creative researchers and advanced engineers with specialized knowledge of a very high level. This graduate school was started in 1953 under the new university system as the new graduate school for Master's and Doctoral courses. It was created as an organization for advanced education and research that further expands upon the mission of the graduate school under the old system. Initially it offered 11 degree fields: mechanical engineering, applied chemistry, fermentation engineering, metallurgy, naval architecture, electrical engineering, precision mechanics, applied physics, communication systems engineering, welding engineering, and structural engineering. In response to advances in science and technology and the demands of society, the school grew substantially and in 1993 had expanded to 22 degree fields. Department of Adaptive Machine Systems The Department of Adaptive Machine Systems has a program that uses integrated education and research to develop advanced products and systems with adaptive intelligence and functions. The program takes a multidisciplinary approach, integrating the areas of materials science, device engineering, manufacturing science, mechanical engineering and robotics, developing students' ability to identify and solve problems relating to the needs of society, and building students' intercultural communication skills. 9

10 Tamim Asfour, Karlsruhe Institute of Technology (KIT) Tamim Asfour is senior researcher at the Karlsruhe Institute of Technology (KIT), Institute for Anthropomatics, Humanoids and Intelligence Systems Lab. He received his diploma degree in Electrical Engineering in 1994 and his PhD in Computer Science in 2003 from the University of Karlsruhe. Since 2003 he is leader of the Humanoid Research Group at the Institute for Anthropomatics. In 2003 he was awarded with the FZI price for his outstanding Ph.D. thesis on sensorimotor control in humanoid robotics and the development of the humanoid robot ARMAR. He is leading the development team of the ARMAR robots and the system integration tasks in the Collaborative Research Center on Humanoid Robots (SFB 588), which is funded by the German Research Foundation (DFG). He is involved in several European Cognitive Systems projects PACO-PLUS, GRASP and InterLink. His major research interest is humanoid robotics. Karlsruhe Institute of Technology (KIT) The Karlsruhe Institute of Technology (KIT) was founded on October 01, 2009 by a merger of Forschungszentrum Karlsruhe and Universität Karlsruhe. KIT bundles the missions of the two institutions: A university of the state of Baden-Wuerttemberg (Universität Karlsruhe) with teaching and research tasks and a large-scale research institution of the Helmholtz Association (Forschungszentrum Karlsruhe) conducting program-oriented provident research on behalf of the Federal Republic of Germany. Within these missions, KIT is operating along the three strategic fields of research, teaching, and innovation. Michael Herrmann, University of Edinburgh, Institute of Perception, Action and Behaviour J. Michael Herrmann graduated from Leipzig University with a Diploma in Mathematical Physics and a doctoral degree in Informatics. He was a postdoctoral researcher at Niels Bohr Institute, Copenhagen, at RIKEN, Laboratory for Information Representation, Tokyo and at the Max Planck Institute for Fluid Dynamics, Department of Nonlinear Dynamics. In 2002 he became assistant professor at Göttingen University, Institute for Nonlinear Dynamics and is since 2008 faculty member at the Institute for Perception, Action and Behaviour, and member of the Institute for Adaptive and Neural Computation and the Institute of Life Sciences of the University of Edinburgh. He is also principal investigator at the BCCN and BFNT Göttingen and cooperates with Otto Bock HealthCare GmbH on projects in biomechanical modelling, adaptive control, and data analysis. His further interests include Computational Neuroscience, Behavioural Robotics, Information Geometry and Cognitive Ageing Research. 10

11 University of Edinburgh, Institute of Perception, Action and Behaviour The Institute of Perception, Action and Behaviour was formed within the School of Informatics by members of the former Departments of Artificial Intelligence, Cognitive Science and Computer Science in It is focused on activities related to bio-mimetic robotics, computer-based visual perception, dynamic control of the interaction of robotic systems with their environment or each other, computer-based generation of external phenomena, such as images, music or actions, and agent-based interaction with other agents or humans, as in computer games and animation. Jun Morimoto, Department of Brain Robot Interface (BRI), ATR Computational Neuroscience Laboratories, Nara Jun Morimoto is head of Department of Brain Robot Interface (BRI) at ATR Computational Neuroscience Laboratories. He received his Ph.D. in information science from Nara Institute of Science and Technology (NAIST), Nara, Japan, in From 2001 to 2002, he was a postdoctoral fellow at the Robotics Institute, Carnegie Mellon University, Pittsburgh, PA, USA. From 2004 to 2009, he was a researcher in Computational Brain Project, ICORP, JST. He joined ATR in Department of Brain Robot Interface (BRI), ATR Computational Neuroscience Laboratories, Nara ATR Computational Neuroscience Laboratories aims to understand the function of the brain through computational neuroscience and to develop a Brain Machine Interface (BMI) as a basic technology for IT that realizes ICT technology accessible to all and futuristic communication by applying our obtained knowledge. Especially, Department of Brain Robot Interface (BRI) is developing a method to control a robot by using brain activities based on machine learning approaches and control theories. Helge Ritter, University of Bielefeld, Cognition and Robotics Lab Helge Ritter studied physics and mathematics at the Universities of Bayreuth, Heidelberg and Munich. After a Ph.D. in physics at Technical University of Munich in 1988 he visited the Laboratory of Computer Science at Helsinki University of Technology and the Beckman Institute for Advanced Science and Technology at the University of Illinois at Urbana- Champaign. Since 1990 he is head of the Neuroinformatics Group at the Faculty of Technology, Bielefeld University. His main interests are principles of neural computation and 11

12 their application to build intelligent systems, with a major focus on cognitive robotics and manual intelligence in particular. In 1999, Helge Ritter was awarded the SEL Alcatel Research Prize and in 2001 the Leibniz Prize of the German Research Foundation DFG. He is co-founder and Director of the Bielefeld Institute for Cognition and Robotics (CoR-Lab) and coordinator of the Bielefeld Excellence Cluster/ Cognitive Interaction Technology" (CITEC). University of Bielefeld, Cognition and Robotics Lab Helge Ritter is with the Institute for Cognition and Robotics (CoR-Lab) and with the Center for Cognitive Interaction Technology (CITEC), which is a research cluster funded in the framework of the German Excellence Initiative. The vision of the two closely collaborating institutes which are both based at Bielefeld University is the realization of technical systems that can be operated easily and intuitively, with a particular emphasis on humanoid robotics in CoR-Lab, and ranging within CITEC also into issues of intelligent interaction with everyday objects, ambient systems and media. Florentin Wörgötter, University Göttingen, The Bernstein Center for Computational Neuroscience Florentin Wörgötter has studied Biology and Mathematics in Düsseldorf. He received his PhD in 1988 in Essen working experimentally on the visual cortex before he turned to computational issues at the Caltech, USA ( ). After 1990 he was researcher at the University of Bochum concerned with experimental and computational neuroscience of the visual system. Between 2000 and 2005 he had been Professor for Computational Neuroscience at the Psychology Department of the University of Stirling, Scotland where his interests strongly turned towards "Learning in Neurons". Since July 2005 he leads the Department for Computational Neuroscience at the Bernstein Center at the University of Göttingen, Physics III. His main research interest is information processing in closed-loop perception-action systems, which includes aspects of sensory processing, motor control and learning/plasticity. These approaches are tested in walking as well as driving robotic implementations. His group has developed the RunBot a fast and adaptive biped walking robot. University Göttingen, The Bernstein Center for Computational Neuroscience The Bernstein Center for Computational Neuroscience at the University Göttingen brings together several research institutions, like University, Max-Plank Institutes and German Primate Center, focusing the effort of about 20 groups in theoretical and experimental 12

13 neuroscience. It is funded by the German Ministry of Science and is part of the Bernstein Network in which all across Germany neuroscience research is being supported in several centers. These efforts include also novel developments in neuro-technology; e.g. prosthetics, artificial cochlea research, and robotics. This way an internationally unique scientific environment has been created in which collaborations are fostered between the neurosciences and other more technologically oriented fields, from which strong mutual benefits arises. 13