Thursday (6/25)

Thursday, June 25

09:00 - 10:00 | Keynote:

Akio Ishiguro, Research Institute of Electrical Communication, Tohoku University

"Toward Understanding the Inter-limb Coordination Mechanism in Legged Locomotion"

Animals are able to exhibit surprisingly adaptive and resilient behavior in real time under real world constraints. Such movements are achieved via spatiotemporal coordination of a large number of bodily degrees of freedom in response to the environment. Clarifying the control principle underlying this remarkable ability of animals allow us to understand biological systems more deeply as well as to construct truly adaptive robot that could not be realized solely by the conventional robotics methodology. We have so far been investigating the control principles underlying various types of animal locomotion, ranging from amoeboid locomotion to human bipedal locomotion, from the viewpoint of decentralized control. In this talk, I will mainly introduce the latest results of our robotic case study on the interlimb coordination mechanism for quadruped locomotion. I will then show how this model could be applied to understanding the interlimb coordination mechanism in legged locomotion with different number of legs.

10:00 - 10:20 | Break / Refeshments
10:30 - 12:00 | Session 4A

Greg Sawicki, Biomedical Engineering, North Carolina State University
Spring-loading locomotion: Considering muscle-tendon dynamics on the human side of the human-machine interface."
First, I discuss the motivation and basic science behind the design of a passive elastic ankle exoskeleton and novel clutching mechanism that can reduce the musculoskeletal loads on plantarflexor muscles by storage and release of elastic energy in a parallel spring worn about the ankle (i.e. exo-tendon) during human walking. Then, I address a crucial question in the design of wearable assistive devices: How does a mechanical element (motor or spring) acting in parallel with biological muscle-tendon unit influence its underlying ‘tuned’ elastic behavior?  I examine how parallel assistance (e.g. from an exoskeleton) alters muscle-tendon interaction using a simple, forward dynamics (i.e. predictive) model of the ankle plantarflexors during vertical hopping with ‘exo-tendons’. When possible, I compare model estimates of muscle dynamics (i.e. length, velocity trajectories) to human data using functional ultrasound images taken from the triceps surae muscles with and without ‘exo-tendon’ assistance. Finally, I discuss how this simple modeling-experimental framework could be used to (1) optimize mechanical assistance in terms of the user’s energetic benefit, injury risk and adaptation time and (2) elucidate underlying mechanisms that may drive preferred behaviors during human movement. A common theme emerges: More is not always better. That is, there are potential side-effects to exoskeleton designs that maximize reductions in musculoskeletal loading by ‘turning up the dial’ on mechanical assistance.

Joo H. Kim, Mechanical and Aerospace Engineering, New York University
"Energy Expenditure Models of Machines and Animals in Motion"
Energy expenditure—as the consumption of an energy source—of actuated multibody systems, such as machines and animals, is an important performance criterion. However, energy expenditure is complicated to measure and difficult to predict. In this study, a rigorous mathematical model of energy expenditure for a general actuated multibody system is established as a function of state variables, control inputs, and system parameters. The model forms are derived theoretically using the laws of thermodynamics and the principles of multibody system dynamics, while the model parameters are estimated experimentally. The internal energy for chemical reactions (e.g., battery, gasoline, and food) and heat are formulated in terms of the kinematic and dynamic variables of the dynamic system. The work components for actuation (e.g., electric motor, internal combustion engine, and muscle), dissipation, restoration, and reaction are formulated with respect to generalized coordinates, and incorporated into the model. The general models are developed to reliably and accurately evaluate instantaneous energy expenditure without limitations inherent in experimental measurements or other approximation models. Experimental and computational results of robotic and human walking energetics will be used as illustrative examples. In addition, a method of integrating the energy expenditure models, along with balance criteria, into a contact optimization algorithm to determine the optimal compromise between efficiency and stability in wearable robot and prosthetic gait will be discussed. 

Max DonelanBiomedical Physiology & Kinesiology, Simon Fraser University.
"On Size and Time"
Animals of all sizes appear to rely on sensory feedback to move. Whether their movement is to escape from a predator, to avoid a fatal fall, or to perform more mundane tasks, the effectiveness of this sensory feedback is constrained by sensorimotor delays. Here, we present our research determining the scaling of total delay and its component delays, in species spanning the full size range of terrestrial mammals. Using a combination of our own electrophysiological experiments and systematic reviews of the literature, we found that some component delays are short and relatively constant, while others increase with animal size.  The three constant delays – sensing, synaptic, and neuromuscular junction delay – offset sharp increases in nerve conduction delay, while muscle’s electromechanical and force generation delays increase more moderately with animal size.  The sum of the six component delays, termed total delay, increases with animal size in proportion to M0.19 – large animals experience substantially longer delays than smaller animals. Intriguingly, sensorimotor delays appear to be tuned to movement times – total delay is nearly independent of animal size when expressed relative to stance phase duration at physiologically equivalent speeds. That is not to say that delays do not limit performance. At sprinting speeds, for example, total delay approaches stance duration making it difficult for an animal to neurally compensate for a perturbation within the same step. Furthermore, it may be important under some circumstances for large and small animals to respond to a stimulus in the same absolute time, such as when stung by a disease-infected insect. In these instances, large animals are at a considerable disadvantage, with total delays almost fifteen times longer than their smaller counterparts. 


David LeeLife Sciences, University of Nevada, Las Vegas 
"Bipedal walking dynamics of humans and birds: Different solutions from different evolutionary legacies"
Birds and humans are the only striding bipeds living today, so birds provide the only explicit comparisons to our own locomotion. It has been known for several decades that, when normalized to body size, stride lengths and stride frequencies of birds and humans are the same at any given dimensionless speed. Here we confirm these similarities, yet find profound differences in the underlying gait dynamics. Humans reduce collisions across their entire walking speed range, evidenced by a low collision fraction maintained at ~0.4. At the transition to human running, collision fraction changes abruptly and is maintained at ~0.95. In contrast, across the speed range corresponding to human walking, birds show a continuum between walking and running dynamics, with collision fractions increasing linearly from ~0.55 to ~0.95. Likewise, collision angles and, therefore, mechanical costs of transport of birds increase linearly from 0.10 to 0.20, approaching the values for human walking at low speeds and human running at higher speeds. Humans use an innovative heel-strike/toe-off transition strategy to minimize collisions, and thus keep collision fraction and collision angle flat across moderate to fast walking. This observation may be related to key differences in the evolutionary legacies of birds and humans, namely the presence of a heel bone in mammals and the plantigrade posture used by arboreal ancestors of humans.

Patrick Wensing, Biomimetic Robotics Lab, MIT
"Towards a Characterization of Gait Energetics for Electrically Actuated Quadrupeds"
Matching the efficiency of quadrupedal animals in legged machines has long been an elusive goal. Recently, new design paradigms for the MIT Cheetah robots, centered on high-torque-density electric motors, have enabled bounding and galloping energetics that rival the biological realm.  While there is much work illuminating factors governing gait selection in animals, little is known about the relative merits of gaits in electrically-actuated robots. This talk will introduce a new study that aims to characterize gait energetics for these machines. Our initial results show that optimal locomotion strategies may match those employed in large ungulates, where muscle force scaling limitations give rise to marked changes in posture in comparison to smaller quadrupeds. The insights gained show promise for the MIT Cheetah to yet exceed the efficiency of its biological counterpart.

12:00 - 1:00   | Lunch
1:20   - 2:50   | Session 4B

Yong-Lae Park, Robotics Institute, CMU
"Adaptive Motion Sensing and Actuation of The Human Body using Soft Wearable Robotic Devices"
It is very difficult to detect and actuate motions of the human body with traditional robotics techniques since many parts of the human body are composed of soft tissues with three-dimensional curved surfaces.  Although the skeletal structure of the body can be mechanically analyzed, external measurement and assistance are not straightforward due to the soft structure.  Furthermore, since traditional robotics take advantage of rigid structures, such as well-defined linkages and joints, some of the natural degrees of freedom of the human body must be mechanically constrained for monitoring and actively assisting body motions, which can be easily seen in exoskeletons.  To solve this problem, we have been investigating various soft robotics technologies focusing on sensors and actuators made of hyperelastic materials.  In this talk, I will first introduce the recent advances in highly stretchable and flexible artificial skin and muscle technologies that could be directly used for body motion sensing and actuation. Then, I will present soft wearable robotic systems we have been developing by integrating the soft sensors and muscle actuators as examples of assistive and rehabilitation technologies. I will also discuss novel 3-D fabrication processes for multi-material, multi-functional soft structures.

Avik De, School of Engineering and Applied Science, University of Pennsylvania
"Parallel Composition of Templates via Average Anchoring"
Raibert described planar hopping as a composition of three "parts": controlled vertical hopping, controlled forward speed, and controlled body attitude. Such reduced degree of freedom compositions also seem to appear in running animals across several orders of magnitude of scale. Dynamical systems theory can offer a formal representation of such reductions in terms of "anchored templates," respecting which Raibert's empirical synthesis (and the animals’ empirical performance) can be posed as a parallel composition. However, the orthodox notion (attracting invariant submanifold with restriction dynamics conjugate to a template system) has only been formally synthesized in a few isolated instances in engineering (juggling, brachiating, hexapedal running robots, etc.) and formally observed in biology only in similarly limited contexts. A new, relaxed view of the anchoring relation ("average anchoring"), offers an analytical framework that has produced a growing collection of (body, behavior) pairs that successfully anchor the three parts of planar hopping using decoupled controllers. We can now show that this relaxation is both necessary and sufficient for the Raibert hopper to have anchored its three templates. The talk will introduce this new formalism and illustrate its empirical value through a variety of such provably correct relaxed compositions on a new tailed monopedal robot platform.

Mirko Kovac, Imperial College London
"Bioinspired Aerial-Aquatic Mobility for Miniature Robots"
Multi-modal locomotion has recently gained attention in the robotics community with several groups demonstrating air-ground and water-ground mobility concepts and working robot prototypes. Aerial-aquatic locomotion however, has been explored only very little despite the vast application opportunities that would be enabled by flying robots that could perform autonomous water sampling and underwater structural assessment. In this talk, I will present the progress of the Aquatic Micro Aerial Vehicles (AquaMAV) research area that is undertaken at Imperial College London where we develop novel, biologically inspired robots that are capable of moving in air, under water as well as across fluid interfaces. 


Robin Thandiackal, Biorobotics Lab, EPFL
A new Type of Salamander-like Robot to Study Various Aspects of Limb Coordination"
A key element to properly understand animal locomotion is to model the interplay between various components such as the neural control, the musculoskeletal system and physical interactions of body and environment. Therefore, researchers have built robots, which connect the numerous models to a physical body or simulated physical bodies. We present the salamander-like robot Pleurobot that was built with a new design methodology for robots based on detailed biplanar high-speed cineradiography. In further studies this robot will allow us to study a rich variety of gaits such as aquatic stepping gaits and stepping gaits on uneven terrain.


Takuya Umedachi, Tufts Univerisity
"Sensing Dynamic Properties and Interactions with the Environment: Proprioceptive Information for Soft-bodied Machines"
This paper presents a sensing method to measure the dynamic properties of a soft-bodied robot and its interaction with the environment by rhythmically activating (bending) a body part with a virtual muscle (which consist of a DC motor, encoder, and wire). Experimental results clearly show that the sensing method is able to distinguish two different bodies and environmental changes exploiting deformation and motion of the soft body. This method is very easy to implement for all soft-bodied robot driven by motors and wires.

3:00               | Farewell Remarks