ICRA 2012 Workshop Variable Stiffness Actuators moving the Robots of Tomorrow St. Paul

ICRA 2012 Workshop

Variable Stiffness Actuators moving the Robots of Tomorrow

St. Paul - Minnesota, USA

May 14, 2012

Abstract

In emerging robotic applications, in which robots need to cooperate closely with humans, traditional rigid structures and stiff actuation systems should be properly modeled and controlled in order to achieve adaptability, compliancy and safety. An alternative design approach is to build actuators with physically adjustable compliance and damping, able to store and release mechanical energy, to react softly when touching the environment, and intrinsically safe, just like human muscles do. Robots and biomechatronic systems can be given similar capabilities as humans by implementing variable impedance actuators, characterized by the property that their apparent output stiffness, and thus the stiffness of the actuated joint, can be changed independently from the actuator output position. Variable impedance actuators realize robust, energy-efficient and highly dynamic systems, which permit the embodiment of natural characteristics, found in biological systems, and implementable into a new generation of robots. This full-day workshop covers the challenges related to the technological realization and the functionalities of such actuation systems, in terms of both mechanical design and control, by mostly focusing on complex robotic and biomechatronic applications, such as robotic manipulation, bipedal and multi-legged locomotion, prosthetics (hands, arms and legs), rehabilitation devices and social robots (humanoids).

List of Topics

Mechanical design of variable impedance actuators, Motion and stiffness control for variable impedance actuators, Mechanical-control co-design, Safety analysis and injury indices in human-robot interaction, Mechanical design of complex robotic and biomechatronic systems based on variable impedance actuations systems (hands, arms, legs, etc.), Control strategies for highly dynamic behaviors of complex systems based on variable impedance actuations systems (throwing, running, drumming, etc.), Impedance observers, Standardization of the description of variable impedance actuators (data sheets), Energy efficiency, Robustness.

Program of the Workshop

• 08:30 - Welcome

• 08:40 - Gill Pratt, DARPA / Defense Sciences Office Challenges in actuation
• 09:20 - Alin Albu-Schäffer, Deutschen Zentrums für Luft- und Raumfahrt VIA systems: from design to control

• 10:00 - Coffee break

• 10:30 - Clément Gosselin, Université Laval Series clutch actuators: a zero-stiffness approach for the design of safe robots
• 11:10 - Nikolaos G. Tsagarakis, Italian Institute of Technology Variable stiffness actuation, energy efficiency issues and inherent damping effects
• 11:50 - Neville Hogan, Massachusetts Institute of Technology Towards a general theory of actuators

• 12:30 - Lunch

• 13:30 - Etienne Burdet, Imperial College Robots can learn to deal with unknown and varying dynamics the human way
• 14:10 - Antonio Bicchi, University of Pisa Variable stiffness actuation, optimal control, and snowclones
• 14:50 - Jonathan W. Hurst, Oregon State University Variable stiffness in legged locomotion
• 15:30 - Mojtaba Azadi, Massachusetts Institute of Technology Variable stiffness joint for locomotion
• 15:45 - Ercan Elibol, University of South Florida Experimental studies on variable motor stiffness during humanoid robot walking

• 16:00 - Coffee break

• 16:30 - Raffaella Carloni, University of Twente Energy efficient and robust variable stiffness actuators
• 16:50 - Bram Vanderborght, Vrije Universiteit Brussels Variable stiffness actuators for human-robot interaction
• 17:10 - Ashish D. Deshpande, University of Texas Toward the development of a robotic hand with programmable passive dynamics
• 17:25 - Marko B. Popovic, Worcester Polytechnic Institute Soft robotics variable stiffness exo-musculature, one-to-many concept, and advanced clutches (Presented by Thane R. Hunt and Christopher J. Berthelette)

• 18:00 - Discussion and Conclusion

List of Presenters

• Alin Albu-Schäffer, Deutschen Zentrums für Luft- und Raumfahrt, Germany

VIA systems: from design to control

• Antonio Bicchi, University of Pisa, Italy

Variable stiffness actuation, optimal control, and snowclones

There are at least two fundamental reasons why Optimal Control is a fundamental tool in studying VSA systems. On one side, it provides a principled basis to compare the performance of different system designs, hence it supports the core of the scientific investigation on the fundamentals of variable impedance systems. On the other side, Optimal Control represents a level of abstraction sufficiently high, but still transparently operational, at which natural and artificial movement science and technology can meet and exchange results. In the talk, we will briefly expand on these two points.

• Etienne Burdet, Imperial College, United Kingdom

Robots can learn to deal with unknown and varying dynamics the human way

Variable impedance actuators can produce arbitrary force and impedance, but how to select suitable values for interacting with unknown or varying environments? Similarly, while humans have muscles that they can coordinate to produce arbitrary force and impedance, they are not able to just select an impedance value. However, humans perform tasks with appropriate force and impedance, which they have learned in repeated movements. This presentation will examine how this learning takes place, and illustrate how robots can learn interacting with unknown and unstable dynamics in a similar way, using simple algorithms requiring little computation and memory.

•Raffaella Carloni, University of Twente, The Netherlands

Energy efficient and robust variable stiffness actuators

• Clément Gosselin, Université Laval, Canada

Series clutch actuators: a zero-stiffness approach for the design of safe robots

As robots are making their way from protected industrial environments to the real world, ensuring the safety of humans interacting with them is a primary concern. Although robust control schemes and high-level behaviour programming are essential, safety ultimately depends on the mechanical design of robots. In this presentation, the concept of series clutch actuators (SCAs) is put forward as a potential candidate for providing inherent mechanical safety. Series clutch actuators are torque limiters placed in series with the robot's actuators and whose limit torque can be adjusted according to the robot's configuration. By properly adjusting the limit torques of the SCAs, the maximum static force that the robot can apply to its environment at the tool centre point (TCP) can be limited to a prescribed safe level. If a limit torque is exceeded, the corresponding SCA slips while maintaining the maximum torque. This zero-stiffness behaviour physically disconnects the inertia of the robot components located upstream from the SCA in the kinematic chain, thereby reducing considerably the dangerousness of the robot. In this presentation, the concept of SCAs is first reviewed. Then, a method is presented to determine the optimal limit torques that maximize the isotropically achievable force (the force that can be applied in all directions without triggering any of the SCAs) while satisfying the prescribed safe force limit. An approach to optimize the pose of a redundant robot in order to maximize the isotropically achievable force while ensuring a safe maximum force threshold is also proposed. Then, the design and fabrication of torque limiters is addressed. Finally, experimental results obtained with a 4-DOF redundant serial arm are presented and discussed.

• Neville Hogan, Massachusetts Institute of Technology, USA

Towards a general theory of actuators

A general mathematical representation of variable-impedance actuators would be useful to facilitate comparison and evaluation of alternative designs. This presentation will propose general features of such a theory in broad outline. Actuators necessarily reside at the interface between the domains of information processing (computation and control) and energy processing (physical system dynamics). The constraints encountered in these domains differ, a fact that may be used to advantage (which will be reviewed by selected examples). Nonetheless, a combined representation of informational and physical dynamics is required. The applicability of equivalent networks for this purpose will be reviewed and explored. Representations of variable-impedance actuators are necessarily nonlinear. With suitable generalization, nonlinear equivalent networks can be defined. Care is required to ensure that the network dynamic elements are unambiguously identifiable. Moreover, the result is not a minimal state-determined representation. However, this representation clearly distinguishes between informational and physical dynamics. A simplified representation of variable-impedance actuators may be obtained by Taylor-series expansion. The result is a multiplicative (bi-linear) form loosely analogous to a representation that has proven effective hitherto for constant-impedance systems. These representations will be illustrated by application to the mammalian neuro-muscular actuator.

• Jonathan W. Hurst, Oregon State University, USA

Variable stiffness in legged locomotion

Biomechanics researchers often describe changes in leg stiffness based on observations of animal running. Animals adjust leg stiffness with speed, accommodate ground impedance changes by changing leg stiffness, and change hop-in-place frequency by changing leg stiffness. We show the same behavior in simulation and demonstration on a hopping robot, and discuss the implementation details.

For all physical interaction tasks, it is clear that careful specification of compliance and other passive dynamic behavior is critical to performance. This talk will also address the question of how to choose the series stiffness of an actuator to maximize bandwidth for a task, based only on limitations of passive dynamics, assuming perfectly optimal control.

• Gill Pratt, DARPA / Defense Sciences Office, USA

Challenges in actuation

• Nikolaos G. Tsagarakis, Italian Institute of Technology, Italy

Variable stiffness actuation, energy efficiency issues and inherent damping effects

Peak power motion generation, adaptability and robustness to interaction, and energy efficiency are some of the performance enhancements that variable impedance actuation can bring to the next generation of robots. Although the first two have been recently successfully demonstrated the use of variable impedance actuation for energy efficiency has not yet been explored. The first part of the talk will discuss on the energy efficiency of VSA joints under periodic motions of fixed and variable frequency while the second part will highlight the benefits gained by the incorporation of intrinsic variable damping in elastic joints.

• Bram Vanderborght, Vrije Universiteit Brussels, Belgium

Variable stiffness actuators for human-robot interaction

The main research topics of the Robotics & Multibody Mechanics (R&MM) research group can be divided into physical human-robot interaction (pHRI) and cognitive HRI (cHRI). Core technology is the design and control of variable stiffness actuators for embodied intelligence principles as safety, energy efficiency and adaptability. The presentation will provide an overview of the different developed systems.

Abstract Submission

We invite submission of abstracts (one A4, in pdf).

Authors of accepted abstracts will be given the opportunity to present their work during the poster session, preceded with a two-minutes presentation describing their poster.

Please email your abstract to Raffaella Carloni (r.carloni@utwente.nl) by 8 April, 2012.

Important Dates