Design of Load-Bearing Exoskeletons for Versatile, High-Mobility Applications

Dr. Nicholas Paine
Apptronik

Abstract: Robots that achieve mobility using legs instead of wheels or tracks offer advantages in versatility at the cost of numerous system design and control challenges.  A subset of legged robots, exoskeletons pose additional considerations tied primarily to the human-robot interface.  While recent advances in this space show promise, many technical obstacles remain.  Current full-body exoskeletons in the industrial and security space are either actively actuated but big, heavy, and bulky, or passively actuated and only useful for a limited set of tasks.  In this talk we will discuss recent work performed in the space of full-body load-bearing exoskeletons intended for able-bodied users.  Our objective is to retain the benefits in versatility of active systems, while significantly improving their drawbacks such as system bulk and mass, power consumption, and overall system transparency to the user.

Bio: Nicholas Paine is co-founder and CEO of Apptronik.  He received his B.S., M.S. and Ph.D. degrees in Electrical and Computer Engineering from The University of Texas at Austin.  His graduate work focused on the development of the UT Series Elastic Actuator, a compact high performance actuator for robotics.  He was a member of the NASA-JSC DARPA Robotics Challenge team where he helped design SEAs and developed actuator-level controllers for the Valkyrie robot.  He worked for one year as a post-doctoral researcher at UT Austin, investigating forced-convective cooling of electric motors and embedded system design.  Since Apptronik’s founding in 2016, he has worked to direct and coordinate both technical and business strategy and execution.  He has helped to lead numerous system design efforts ranging from dynamic balancing humanoids, to force-augmentative exoskeletons and novel robotic manipulators.

What’s needed for useful walking robots?

Dr. Robert Griffin
IHMC

Abstract: Walking robots, from bipeds to quadrupeds to exoskeletons, have progressively gotten better in recent years, with quadrupeds now reliably going off-road and humanoids capable of highly dynamic motions. This raises the question of what skills are needed for humanoid robots to make the transition from research platforms to robots that can be used in the real world? In this talk, we will explore this question, and highlight some of the recent research at IHMC that we are undertaking in hopes of find an answer.

Bio: Dr. Robert Griffin is a Research Scientist at the Florida Institute for Human and Machine Cognition (IHMC). He received his B.S. from Tennessee Technological University and his Ph.D. from Virginia Tech, joining IHMC in 2016. His work primarily focuses on planning and control of robotic systems, particularly focusing on legged robots, including exoskeletons, humanoids, and quadrupedal robots. Since joining IHMC, he has been developing algorithms and assisting in managing the IHMC robotics team, currently leading a number of the projects at IHMC. This includes the Office of Naval Research’s Squadbot project, which aims to develop a new, dynamic, high range of motion humanoid robot, as well as ONR’s High Speed Humanoid Behaviors project, with the goal of developing autonomous and semi-autonomous behaviors for humanoid robots functioning in urban environments. He also leads IHMC’s exoskeleton efforts, currently focused on developing lower-body exoskeletons for people with paralysis as part of the 2020 Toyota Mobility Unlimited Challenge and the 2020 Cybathlon. He is also a key researcher in the NASA Johnson Space Center’s Val-EOD project, where the goal is to develop humanoid robots, primarily NASA JSC’s Valkyrie, to function as Explosive Ordnance Disposal operators. He also led IHMC’s effort in the Army Research Laboratory’s Robotics Collaborative Technology Alliance, developing a full-scale quadruped control stack for autonomous planning and locomotion over rough terrain. He was also the lead software and controls developer for IHMC’s exoskeleton entrant into the 2016 Cybathlon, where they placed second. Prior to joining IHMC, Dr. Griffin was a key controls engineer for the Virginia Tech DARPA Robotics Challenge team, as well as the ONR Shipboard Autonomous Firefighting Robot program.

From kinematic to energetic design and control of wearable robots for agile human locomotion

Prof. Robert Gregg
University of Michigan

Abstract: Even with the help of modern prosthetic and orthotic (P&O) devices, lower-limb amputees and stroke survivors often struggle to walk in the home and community. Emerging powered P&O devices could actively assist patients to enable greater mobility, but these devices are currently designed to produce a small set of pre-defined motions. Finite state machines are typically used to switch controllers between discrete phases of the gait cycle, e.g., heel contact vs. toe contact, and between different tasks, e.g., uphill vs. downhill. However, this discrete methodology cannot continuously synchronize the robot’s motion to the timing or activity of the human user. This talk will first present a continuous parameterization of human joint kinematics based on a phase variable that robustly represents the timing of the human gait cycle. This parameterization is employed for user-synchronized control of a powered knee-ankle prosthesis, which enables above-knee amputee subjects to walk at variable speeds with reduced compensations of intact joints. This method also allows the user to perform volitional activities like kicking a ball and stepping over obstacles. To fully leverage this control approach, this talk will introduce a quasi-direct drive approach to actuating prosthetic legs for accurate control of joint torque and impedance, enabling more dynamic motions, reducing power consumption, and reducing acoustic noise compared to state-of-art robotic prostheses. While these methods reproduce missing joint function, a different control philosophy is needed for exoskeletons that assist existing joint function. The last part of this talk will introduce an energetic control paradigm for exoskeletons to alter the human body’s dynamics without prescribing joint kinematics, e.g., reducing the perceived weight and inertia of the limbs. This control approach is implemented on exoskeletons with high-torque, low-impedance actuators, which provide the necessary backdrivability for volitional human control.

Bio: Robert D. Gregg IV received the B.S. degree in electrical engineering and computer sciences from the University of California, Berkeley in 2006 and the M.S. and Ph.D. degrees in electrical and computer engineering from the University of Illinois at Urbana-Champaign in 2007 and 2010, respectively. He joined the University of Michigan as an Associate Professor in the Department of Electrical Engineering and Computer Science and the Robotics Institute in Fall 2019, and he became Associate Director of Robotics in Fall 2020. Prior to joining U-M, he was an Assistant Professor in the Departments of Bioengineering and Mechanical Engineering at the University of Texas at Dallas. Dr. Gregg directs the Locomotor Control Systems Laboratory, which conducts research on the control mechanisms of bipedal locomotion with applications to wearable and autonomous robots. He is a recipient of the Eugene McDermott Endowed Professorship, NSF CAREER Award, NIH Director’s New Innovator Award, and Burroughs Wellcome Fund Career Award at the Scientific Interface

Beyond Inverted Pendulums: Optimizing Task-driven Simple Models

Prof. Michael Posa
UPenn

Abstract: Recent advances in hardware development have made dynamic, high-dimensional robots like the Boston Dynamics Spot, the Agility Robotics Cassie, the Ghost Robotics Minitaur, and many others widely accessible. To achieve real-time control and planning of these complex systems, the community has typically relied upon leverage low-dimensional models which capture critical aspects of the full system. For legged locomotion, these models have typically taken the form of inverted pendulums or centroidal masses, and have enabled much of the recent progress on walking and running robots. At the same time, these methods cannot represent the full capabilities of the original robots; this inevitability has led to a large set of hand-engineered extensions and alternatives. In this talk, I will present our recent work to automatically synthesize simple models designed to succeed across a space of tasks. Our approach leverages trajectory optimization within bilevel optimization to improve upon traditional models, which we can then deploy for real-time, high-performance planning. Time-permitting, I will also discuss some of our recent work on leveraging representations of non-smooth dynamics to greatly improve the performance of machine learning when applied to multi-contact robotics.

Bio: Michael Posa is an Assistant Professor in Mechanical Engineering and Applied Mechanics at the University of Pennsylvania. He leads the Dynamic Autonomy and Intelligent Robotics (DAIR) lab, a group within the Penn GRASP laboratory.  His group focuses on developing computationally tractable algorithms to enable robots to operate both dynamically and safely as they quickly maneuver through and interact with their environments, with applications including legged locomotion and autonomous manipulation. Michael received his Ph.D. in Electrical Engineering and Computer Science from MIT in 2017, where, among his other research, he spent time on the MIT DARPA Robotics Challenge team. He received his B.S. in Mechanical Engineering from Stanford University in 2007. Before his doctoral studies, he worked as an engineer at Vecna Robotics in Cambridge, Massachusetts, designing control algorithms for the BEAR humanoid robot. He received the Best Paper award at Hybrid Systems: Computation and Control in 2013, was a finalist for Best Oral Paper at Humanoids in 2016, and received a Google Faculty Research Award in 2018.

Post-failure Actions in Humanoids

Dr. Jinoh Lee
DLR

Abstract: The need for robots to cope with unstructured environments and replace humans in hazardous tasks became an important virtue. Humanoids are naturally highlighted for their potential capability to access to unstructured and uncertain environments. However, one of the significant barriers to practically operate the robot is the difficulty to avoid failure situation such as fall-over and faults in actuation systems. The robots are compulsory to sustain minimal damages and autonomously recover by itself for seamlessly accomplishing their tasks without human intervention. In this regard, this talk will offer a brief view on the research trend of post-failure controls, and introduce recent control strategies which allow the robot to reactively generate the entire body motion to protect and recover itself after failures.

Bio: Jinoh Lee is currently a Research Scientist with the Institute of Robotics and Mechatronics, German Aerospace Center (DLR), Weßling, Germany. He received the B.Sc. degree in mechanical engineering from Hanyang University, Seoul, South Korea, in 2003 awarded Summa Cum Laude and the M.Sc. and the PhD degrees in Mechanical Engineering from Korea Advanced Institute of Science and Technology (KAIST), Daejeon, South Korea, in 2012. He held the postdoctoral position from 2012-2017 and research scientist position from 2017-2020 at the Department of Advanced Robotics, Istituto Italiano di Tecnologia (IIT), Genoa, Italy. His professional is about robotics and control engineering which includes compliant system control for safe human-robot
interaction, whole-body manipulation of high degrees-of-freedom humanoids, and robust control theories for highly nonlinear systems. He is currently a Senior Member of IEEE and serves as AE of IEEE conferences such as ICRA and IROS; the Technical Committee of International Federation of Automatic Control (IFAC), TC4.3 Robotics.

Motion generation for humanoid and complex robots: from Motion Planning to Whole body controller. 

Dr. Olivier Stasse
CNRS

Abstract: In this talk, I will give an outline of the current pipeline developed in the Gepetto Group to generate motion for humanoid and quadruped robots. The blocks start from  a motion planner, then go through trajectory optimizers which can be used either for model predictive
control or optimal control together with a whole body controller.
Our motion planner is able to compute very quickly a trajectory for a humanoid robot to go from a starting configuration to a goal configuration while making multiple contacts. Differently from previous approaches the planner is able to find trajectories which cannot be
found from a quasi static trajectory,  but only through a dynamical transition between two contacts. The same planner can also be used to plan automatically a sequence of stacks
needed to turn upside down an object. I will also present a new Differential Dynamic Programming solver which is extremly efficient for both the humanoid robot TALOS and the open source quadruped SOLO.

Bio: Olivier Stasse received in 2000 a Ph.D. on Intelligent Systems from the University of Paris 6, 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 at the Joint French-Japanese Robotics Laboratory (JRL) between the CNRS and the AIST in Tsukuba.  In 2011 he joined the Gepetto team at LAAS, Toulouse. His research interest lies in fast decision making to generate motion for humanoid robotics.

News from the Skippy Project: Reaching for the Performance Envelope

Dr. Roy Featherstone
IIT

Abstract: The Skippy project aims to create a highly athletic monopedal robot, called Skippy, having only two actuated degrees of freedom, and yet able to balance and orient itself quickly in 3D on its single point foot, and to perform precise hops and somersaults to a height of 3m. Skippy will be able to crash at high speed without damage, and get up unaided after a fall. Right now, the design of Skippy is nearly complete, and construction will soon begin. This talk will cover some of the ideas that went into the design of Skippy, as well as recent results on the high-performance balance control that Skippy will need.

Bio: Dr. Featherstone is a researcher specializing in robot dynamics and related fields. He is currently a professor in the Department of Advanced Robotics at the Italian Institute of Technology, where he works mainly on the Skippy Project, which aims to create a highly athletic monopedal robot. He previously worked at the Australian National University; and before that he was a lecturer in the Department of Computer Science at the University of Wales, Aberystwyth, an EPSRC Advanced Research Fellow in the Department of Engineering Science (Robotics Group), Oxford University, and a senior researcher at Philips Research Laboratories, Briarcliff Manor, New York. He obtained his Ph.D. at Edinburgh University (Dept. of Artificial Intelligence) and his first degree at Southampton University. He is one of the world’s leading experts on rigid-body dynamics, the inventor of the Articulated-Body
Algorithm and the author of two books on dynamics. He is also a Fellow of the IEEE and a member of its Robotics and Automation Society. His URL is http://royfeatherstone.org

Learning-based controllers for legged robots

Prof. Jemin Hwangbo
KAIST

Abstract: Legged robots pose one of the greatest challenges in robotics. Dynamic and agile maneuvers of animals cannot be imitated by existing methods that are crafted by humans. A compelling alternative is reinforcement learning, which requires minimal craftsmanship and promotes the natural evolution of a control policy. However, so far, reinforcement learning research for legged robots is mainly limited to simulation, and only few and comparably simple examples have been deployed on real systems. The primary reason is that training with real robots, particularly with dynamically balancing systems, is complicated and expensive. Recent algorithmic improvements have made simulation even cheaper and more accurate at the same time. Leveraging such tools to obtain control policies is thus a seemingly promising direction. However, a few simulation-related issues have to be addressed before utilizing them in practice. The biggest obstacle is so-called reality gap — discrepancies between the simulated and the real system. Hand-crafted models often fail to achieve a reasonable accuracy due to the complexities of actuation systems of existing robots. Therefore, this talk will focus on how such obstacles can be overcome. The main approaches are twofold: a fast and accurate algorithm for solving contact dynamics and a data-driven simulation-augmentation method using deep learning.

Bio: Assistant Professor Jemin Hwangbo received his PhD in 2019 from ETH Zurich, Switzerland. After graduation, he continued working at ETH Zurich as a postdoctoral fellow for one more year until he joined KAIST. He is the recipient of the ETH Medal in 2019 for his outstanding PhD thesis. He is actively working on applications of reinforcement learning on robotics. He developed a method of modeling a robotic system using both an analytical model and data. This idea was published to Science Robotics and featured as one of the 10 remarkable papers from 2019 by Nature. He is a developer of RaiSim physics engine, which is used by many researchers worldwide. Using his physics engine, he aims at combining artificial intelligence and the physical world.