Sven BehnkeHead of Computer Science Department VI and Autonomous Intelligent Systems Group, Institute for Computer Science, Universität Bonn
Talk Title
Hybrid Driving-Stepping Locomotion in Challenging Environments Talk Abstract
T.B.D. Short Bio:
Sven Behnke received his MS degree in Computer Science (Dipl.-Inform.) in 1997 from Martin-Luther-Universität Halle-Wittenberg. In 2002, he obtained a PhD in Computer Science (Dr. rer. nat.) from Freie Universität Berlin. He spent the year 2003 as postdoctoral researcher at the International Computer Science Institute, Berkeley, CA. From 2004 to 2008, Professor Behnke headed the Humanoid Robots Group at Albert-Ludwigs-Universität Freiburg. Since April 2008, he is professor for Autonomous Intelligent Systems at the University of Bonn and director of the Institute of Computer Science VI. His research interests include cognitive robotics, computer vision, and machine learning. A list of his program committee memberships, meeting organizations, and invited talks can be found here. |
Johannes EnglsbergerResearch fellow at the Institute for Robotics and Mechatronics, DLR
Talk Title
Simple yet powerful walking gait generation based on the Divergent Component of Motion (DCM) Talk Abstract
In this talk, I will first recapitulate the basics of the Divergent Component of Motion (DCM) and the corresponding 3D force application points (eCMP and VRP; similar to ZMP). I will then present our most recent DCM-based walking trajectory framework. It is purely analytical (everything expressed in matrix form) and thus real-time capable. It generates continuous and consistent DCM, VRP and (if desired) CoM trajectories for a multi foot step preview. In combination with a PD tracking controller (or embedded into an MPC framework), the generated trajectories are a really powerful tool for successful walking. I will give my best to describe the method in full detail such that everyone (even if new to the field) can easily follow the talk. Short BIo
Johannes Englsberger is a research fellow at DLR (German Aerospace Center) at the Institute for Robotics and Mechatronics. He received his PhD from TUM in 2016. His main research interests lie in the field of bipedal locomotion (both walking and running), whole-body control and humanoid robot design. |
Hartmut GeyerAssociate Professor at Carnegie Mellon University's Robotics Institute
Talk Title
Reactive Control of Locomotion in Humans and Powered Prosthetics Talk Abstract
Model-based optimization is the standard tool for generating humanoid locomotion; we reason about the desired center of mass behavior of a humanoid system and then treat its rigid body chain as a black-box constraint toward generating the desired behavior. In the process, we ignore domain knowledge about the (bio)mechanics of segmented legs. The talk will review this knowledge and show how it provides a window into human leg control and an avenue toward reactive leg prostheses. Whether this domain knowledge can add value to control algorithms of humanoids remains open. Potential benefits lie in faster optimizations, more lifelike motions, and more robust behavior. Short Bio
Hartmut Geyer received the Diploma degree in Physics and the Ph.D. degree in Biomechanics from the Friedrich-Schiller-University of Jena in 2001 and 2005, respectively. In 2006, he was awarded an EU Marie Curie Fellowship and worked as a postdoctoral researcher at MIT’s Biomechatronics Group and at the Institute for Automatic Control of ETH Zurich, which he later joined as a postdoctoral associate. He joined Carnegie Mellon’s Robotics Institute in 2010, where he is currently an associate professor. His research focuses on the principles of legged dynamics and control, their relation to human neuromuscular control, and resulting applications in humanoid and rehabilitation robotics. |
Marco HutterAssistant professor for Robotic Systems at ETH Zurich
Talk Title
ANYmal outdoors - autonomous locomotion in realistic terrain Talk Abstract
In this talk, I will present our recent developments in planning and control algorithms that enable the quadruped ANYmal to autonomously locomote in natural environments. I will show how we combined online robocentric elevation mapping with optimization-based foothold planning and posture adaptation to overcome previously unperceived obstacles. Moreover, we give an insight into some of our work on terrain (error) model learning that allows adapting the control as a function of the terrain characteristics. Short Bio
Marco Hutter is assistant professor for Robotic Systems at ETH Zurich, Branco Weiss Fellow and co-founder of ANYbotics. Marco is part of the national competence centers for robotics (NCCR robotics) and digital fabrication (NCCR dfab). His group is participating in several research projects, industrial collaborations, and international competitions that target the application of high-mobile autonomous vehicles in challenging environments such as for search and rescue, industrial inspection, or construction operation. Marco's research interests are in the development of novel machines and actuation concepts together with the underlying control, planning, and optimization algorithms for locomotion and manipulation. |
Sangbae KimAssociate Professor of Mechanical Engineering at MIT
Talk Title
MIT Cheetah : a new design paradigm for physical interaction Talk Abstract
Recent technological advances in legged robots are opening up a new era of mobile robotics. In particular, legged robots have a great potential to help disaster situations or elderly care services. Whereas manufacturing robots are designed for maximum stiffness, allowing for accurate and rapid position tracking without contact, mobile robots have a different set of hardware/software design requirements including dynamic physical interactions with environments. Events such as the Fukushima power plant explosion highlight the need for robots that can traverse various terrains and perform dynamic physical tasks in unpredictable environments, where robots need to possess compliance that allows for impact mitigation as well as high force capability. The talk will discuss the new mobile robot design paradigm focusing on the actuator characteristics and the impulse planning algorithms. As a successful embodiment of such paradigm, the talk will introduce the constituent technologies of the MIT Cheetah 3 capable of jumping over a obstacle, climbing stairs, and run at 3m/s. Short Bio
Prof. Sangbae Kim, is the director of the Biomimetic Robotics Laboratory and an Associate Professor of Mechanical Engineering at MIT. His research focuses on the bio-inspired robot design by extracting principles from animals. Kim's achievements on bio-inspired robot development include the world's first directional adhesive inspired from gecko lizards, and a climbing robot, Stickybot, that utilizes the directional adhesives to climb smooth surfaces featured in TIME's best inventions in 2006. Recent achievement includes the development of the MIT Cheetah capable of stable outdoor running up to 13mph and autonomous jumping over an obstacles at an efficiency of animals. This achievement was covered by more than 300 media articles. He is a recipient of best paper award from International Conference on Robotics and Automation (2007), King-Sun Fu Memorial Transactions on Robotics (2008) and IEEE/ASME transactions on mechatronics (2016), DARPA Young Faculty Award (2013), NSF CAREER award (2014), and Ruth and Joel Spira Award for Distinguished Teaching (2015). |
Joohyung KimResearch Scientist at Disney Research
Talk Title
T.B.D. Talk Abstract
T.B.D. Short Bio
Joohyung Kim is currently a Research Scientist in Disney Research, LA. His research interests include implementation of robots based on animation characters, soft human-robot interaction, balancing/walking control for humanoid robots, and novel mechanisms for legged locomotion. He received BSE and Ph.D. degrees in Electrical Engineering and Computer Science from Seoul National University, Korea, in 2001 and 2012. Prior to joining Disney Research, he was a postdoctoral fellow in Robotics Institute of Carnegie Mellon University for DARPA Robotics Challenge in 2013. From 2009 to 2012 he was a senior engineer in Samsung Electronics, Korea, developing biped walking controllers for humanoid robots. |
Andreea RadulescuPost Doc at Italian Institute of Technology
Talk Title
Robust Locomotion Strategies on the HyQ Robot Series Talk Abstract
The Dynamic Legged Systems Lab at the Italian Institute of Technology focuses on research concerning the design and control aspects of agile legged robots. We are interested in the development of legged robotic systems, chiefly hydraulic quadrupeds, and we are investigating ways that can increase flexibility and performance of legged designs. In our work we underline the need for accurate and robust control, ranging from the level of individual joints up to the overall behaviour of the legged robot. Alongside accurate hydraulic force/torque control at the joint level we are investigating ways of creating and using a variety of different locomotion gaits, that are robust to external disturbances and changing environmental conditions. In this talk I will present an overview of the various locomotion approaches employed on the HyQ robot series, as we extend its deployment from controlled lab environments into real-life applications. Short Bio
Andreea Radulescu is a Postdoctoral Researcher in the Dynamic Legged Systems Group at the Italian Institute of Technology. She received her Ph.D (2016) and Msc (2011) in Intelligent Robotics from the University of Edinburgh, working under the supervision of Prof. Sethu Vijayakumar. Previously, she graduated from the Polytechnic University of Bucharest , with a BSc in Engineering in Automatic Control and Applied Informatics. Her research focuses on optimal control, machine learning and using variable impedance actuators for systems in domains with contacts. She is currently working on locomotion strategies for the HyQ robot series. |
Daniel RixenProfessor of Mechanical Engineering, Technical University of Munich
Talk Title
Versatile and Robust Walking in Uneven Terrains Talk Abstract
In theory, legged robots have conceptional advantages over conventional wheeled robots when navigating in unstructured environments. However, these advantages can only be fully exploited when the control methods of the robot are able to deal with such scenarios in terms of versatile movements and robustness. In this talk, we first present an environment modeling method based on point cloud information of the on-board RGB-D sensor. Then we show our real-time autonomous navigation approach, which uses continuous and graph-based search algorithms. Furthermore, we present our recent advances to robust walking in the presence of disturbances and unknown terrain. This includes the real-time model-based predictive adaptation of foot step positions based on a reduced dynamic model of the robot. In this context, we also briefly present our solution to the mitigation of disturbances in the presence of unknown obstacles, which combines our approaches for autonomous navigation and disturbance rejection. Short Bio
Daniel Rixen received his PhD from the University of Liège (Belgium) in 1997, then spent two years at the Center for Aerospace Structures of the University of Colorado. From 2000-2012, he was head of the chair for Engineering Dynamics at the Technical University of Delft (The Netherlands) and since 2012 he directs the chair for Applied Mechanics of the Technical University of Munich (Germany). He authored and co-authored several books and articles on the topics of numerical modeling for structural and multiphysical systems, experimental dynamics and mechatronics. |
Olivier StasseSenior CNRS researcher at LAAS, Toulouse, France.
Talk Title
Using a Memory of Motion to Efficiently Warm-Start a Nonlinear Predictive Controller Talk Abstract
First I will briefly present the results of the Gepetto team on tackling planning and control problem to realize generalized locomotion on humanoid robots. Then a spotlight will be given in our new approach on using a memory of motion to efficiently warm-start a nonlinear predictive controller. Predictive control is an efficient model-based methodology to control complex dynamical systems. In general, it boils down to the resolution at each control cycle of a large nonlinear optimization problem. A critical issue is then to provide a good guess to initialize the nonlinear solver so as to speed up convergence. This is particularly important when disturbances or changes in the environment prevent the use of the trajectory computed at the previous control cycle as initial guess. In this paper, we introduce an original and very efficient solution to automatically build this initial guess. We propose to rely on off-line computation to build an approximation of the optimal trajectories, that can be used on-line to initialize the predictive controller. To that end, we combined the use of sampling-based planning, policy learning with generic representations (such as neural networks), and direct optimal control. We first propose an algorithm to simultaneously build a kinodynamic probabilistic roadmap (PRM) and approximate value function and control policy. This algorithm quickly converges toward an approximation of the optimal state-control trajectories (along with an optimal PRM). Then, we propose two methods to store the optimal trajectories and use them to initialize the predictive controller. We experimentally show that directly storing the state-control trajectories leads the predictive controller to quickly converges (2 to 5 iterations) toward the (global) optimal solution. The results are validated in simulation with an unmanned aerial vehicle (UAV) and other dynamical systems. Short Bio
In 2000, he received a Ph.D. in Intelligent Systems from the University of Paris 6 under the supervision of P. Coiffet, and the French Habilitation to Supervise Research (HDR) in Robotics (2013) from the University of Toulouse III. From 2000 to 2003, he was assistant professor at the Univ. of Paris XIII. From 2003 to 2011, he was one of the funding members of the Joint French-Japanese Robotics Laboratory (JRL) between the CNRS and the AIST in Tsukuba, Japan and created by P. Coiffet and K.Tanie. |
Tomomichi SugiharaAssociate Professor, Department of Adaptive Machine Systems, Graduate school of Engineering, Osaka University
Talk Title
Subsumption Architecture For Flexible and Robust Biped Locomotion Talk Abstract
In this talk, how to design the locomotion system of biped robots that can move flexibly and robustly in dynamic worlds is discussed. The robot should perceive how the world and the robot itself are and determine how to behave in real-time even from uncertain, incomplete, noisy and unexpected sensory information. The information should be processed continuously and dynamically in both perception and action. Subsumption architecture matches the idea, each subsystem of which should be carefully designed. Some key technologies to build up such a system are presented. Short Bio
Tomomichi Sugihara is an associate professor at Department of Adaptive Machine Systems, Graduate School of Engineering, Osaka University. He received his PhD from the University of Tokyo in 2004. He was an academic research assistant from 2004 to 2005 at the University of Tokyo, and became a research associate. He worked at Kyushu University as a guest associate professor from 2007 to 2010. He moved to Osaka University in 2010 and held the current position. His research interests include kinematics and dynamics computation, motion planning, control, hardware design, and software development of anthropomorphic robots. He also studies human motor control based on robotic technologies. |
Alexander WinklerPh.D. Student at ETH Zurich
Talk Title
Dynamic motion generation over non-flat terrain for monopeds, bipeds and quadrupeds Talk Abstract
Trajectory Optimization is attractive, because once the physics of the problem have been properly modeled, the algorithm would, in an ideal case produce motions for any high-level task, solving legged locomotion planning on a general level. In this talk, I will present out latest efforts in this direction with our open-source algorithm TOWR that simultaneously determines the gait sequence, step-timings, footholds, swing-leg motions and 6D body motion over non-flat terrain. I will explain how the optimization problem is able to efficiently optimize over the discrete gait sequence, while still keeping the variables continuous. Finally, I will show how we transferred some of these motions to our quadruped ANYmal and discuss what might be missing to push the performance of real systems further. (Video) Short Bio
Alexaner W. Winkler is a PhD student at the Agile and Dexterous Robotics Lab and the Robotic Systems Lab at ETH Zurich. He received his Bachelors and Masters (with distinction) in Mechanical Engineering from the Karlsruhe Institute of Technology (KIT) in 2012 and 2013. Before starting his PhD in 2014, he was a researcher at the Dynamic Legged Systems Lab at the Italian Institute of Technology (IIT). His research focuses on the optimal planning and control of dynamic motions for legged systems. |
Chengxu ZhouPost Doc at Italian Institute of Technology
Talk Title
Humanoid Balancing and Its Benchmarking Talk Abstract
T.B.D. Short Bio
Chengxu Zhou is a Postdoc of Humanoid and Human Centered Mechatronics Research Line at Italian Institute of Technology. He received his Bachelor’s degree in mechanical engineering and automation in 2007 from Northeastern University, China, Master’s degree in mechanical engineering from Yamaguchi University, Japan in 2010, and Ph.D. in robotics from University of Genoa. His research interest lies in the dynamic motion control with current focus on the dynamic walking and balancing of humanoid robots, as well as the state estimation and optimization based motion control/planning of legged robots. |