Wednesday (6/24)

Wednesday, June 24

09:00 - 10:00 | Keynote

  Daniel Goldman, School of Physics, Georgia Institute of Technology

"Swimming lizards, sidewinding snakes, and digging ants: animal and robophysical experiments reveal principles of effective environmental interaction"

Using three examples, I will discuss how animal experiments coupled to ``robophysical” modeling enables advances in biology and soft matter physics, and creates more life-like robots. First, 1) to illustrate how this approach can lead to new biological control templates and soft matter physics (when models of the environment are not available), I will discuss our studies of subsurface sand-swimming. We discovered that the sand-swimming sandfish lizard [Maladen et al, Science, 2009] uses a travelling wave of body undulation to swim through dry granular media. Robot experiments and computational models have enabled us to understand how the animal controls its gait to maximize speed and minimize energy use during locomotion. These models have also given us confidence to develop a new class of continuum models (based on resistive force theory) which show promise for theoretical explanation of locomotion in complex environments. Next 2) to demonstrate how the discovery of basic locomotor principles can advance real-word robots, I will next discuss our studies of the locomotion of sidewinding snakes on dry granular media. I will show how, based on animal experiments performed at Zoo Atlanta (in collaboration with Dr. Joseph Mendelon III), we have used a multi-module robot (in collaboration with Prof. Howie Choset’s group at Carnegie Mellon) to reveal how the snakes modulate orthogonal body waves (control templates) to manipulate the substrate (e.g. remain below the yield stress) to climb sandy slopes and perform turning maneuvers. Finally, 3) to illustrate the benefits of this approach in multi-agent systems, I will discuss collective soil excavation by social organisms (fire a­nts) in crowded and confined conditions. Theoretical models inspired by measurements of colony members’ activity during excavation of soil “pellets” in narrow tunnels indicate that the digging fire ant workload distribution enables high performance in crowded, confined conditions. We test the theory using fully autonomous digging robot models. An aggressive digging strategy yields benefits as worker number increases (in terms of the rate of tunnel growth and energy use) but these benefits diminish when the number of workers exceeds a certain value; the biological workload distribution buffers excavation efficacy against crowding.  In summary, I emphasize the need for detailed systematic laboratory robot experiments (what we call robophysics) to provide models for biology as well as to improve real-world robot performance.

10:00 - 10:20 | Break / Refeshments
10:20 - 12:00 | Session 3A

Howe Choset, Robotics Institute, Carnegie Mellon University
"The shape of whole-body locomotion"
We have been investigating locomotion strategies for whole-body systems with a focus on snake robots and other systems whose investigation supports snake robot locomotion. We originally took a fundamental approach to define a space expressive enough to capture the motion of whole-body systems (both biological and robotic), yet concise enough to allow a few parameters to intuitively represent this space in a way that is tractable to analysis and optimization. Our investigation led us to consider the space of shapes which forms the basis for whole-body systems; a closed loop in this space forms a gait.  We defined a new way to derive the functional relationship between a gait and its net displacement with analytic guarantees. Our early work assumed that the shape space can be parameterized with a small set of basis functions, sometimes as small as two.  We then turned to biological systems to develop a data driven approach that extracted basis functions on which biologically inspired gaits can be prescribed.  Leveraging these results, we developed  a novel model-based optimization approach used to simultaneously derive optimal basis functions and optimal gaits that enable new more efficient motion of snake robots.  Allowing biology to guide and geometric methods to ground our search for new motions of the robotic system has in turn led us to ask, and in some cases answer, deeper and more meaningful questions about the whole-body motion of biological systems with our esteemed colleagues from GA Tech.

Craig McGowan, Biological Sciences, University of Idaho
"Bouncing Without Springs"
Bouncing gaits such as running, trotting and hopping are believed to be efficient because they enable animals to utilize elastic energy storage and recovery, much like a bouncing ball. However not all bipedal hoppers have compliant tendons that can serve as springs. Although considerably smaller, kangaroo rats use a bipedal hopping gait that is very similar to that of kangaroos and wallabies. Yet their ankle extensor tendons are relatively thick and stretch very little during hopping. Here we explore potential functional trade-offs between hopping with and without compliant tendons. In a series of experiments with kangaroo rats, we examine whole limb, joint, and individual muscle mechanics in response to changes in mechanical demand. The results of these studies show that while at the joint level kangaroo rats maintain spring-like behavior, this does not reflect the mechanical output of the underlying muscle-tendon units. Rather, dynamic coupling of the joints via biarticular muscles enables muscles to behave as pure motors or dampers. This linkage also enables power developed by proximal muscles to be delivered to the environment via distal joints.  

Brook Flammang, Biological Science, New Jersey Institute of Technology
"A bioinspired long-term, reversible underwater adhesive mechanism"
Adhesion, and in particular long-term reversible adhesion, to a wet or submerged surface is challenging. In the natural world, few organisms can adhere to underwater substrates and those that do generally use glue-like mechanisms or attach only to stationary objects. Remora fishes have evolved a unique adaptive ability - an adhesive disk formed from dorsal fin elements – that allows them to attach reversibly to actively deforming bodies of varying roughness that move at high speed. Our primary objective is to characterize the disc tissues and functional structures via biological and engineering methods. We are accomplishing this through the simultaneous study of live remora, mathematical modeling of functional parameters, and physical modeling of mechanical actions. Five areas of our ongoing research target the multiple functional components of the remora adhesive mechanism: disc morphology, lamellar kinematics, spinule friction, viscoelastic seal performance, and drag hydrodynamics. We are applying our findings to produce a robotic device capable to sticking to surfaces of varying roughness in air or water.

Barry Trimmer, Biology, Tufts Universiry
"Climbing in Co
mplex Environments: Robots und Raupen"

Climbing is a particularly complex form of locomotion requiring animals to support their own weight in both compression and tension with the ability to grasp and release the substrate and still maintain stability. Most climbing robots are designed to operate on flat surfaces or by wrapping around poles. Extending maneuverability into complex three-dimensional structures generally requires additional gripping systems such as suckers or hooks. Caterpillars (Raupen) are an excellent model system for understanding how this might be achieved using a compliant passive gripping system called the proleg. Each of these soft fleshy lobes can passively deploy a semicircular crown of scimitar-shaped hooks that are actively released without dragging or producing any resistance to lift. The most remarkable part of this gripping system is that the hooks can be rapidly detached through the activation of a single muscle regardless of the substrate roughness or the angle of attachment. We have studied the system in detail from both a neural and mechanical perspective. We find that, during climbing, the timing of neural activation of locomotory muscles is changed relative to the release of grip. These results have implications for the design of a simple robotic gripping system that can be turned on or off with a single actuator. Although developed for robotic applications this gripping system could be adapted for use as a wearable device to make human climbing safer.


Takeshi Kano, Research Institute of Electrical Communication, Tohoku University 
"TEGOTAE-based control for one-dimensional crawling locomotion"
Autonomous decentralized control is a key concept for developing robots that move adaptively and resiliently like real animals. However, systematic method for designing it has not yet been established. To tackle this problem, we focused on ``TEGOTAE", a Japanese word meaning reaction after the generation of some action. We adopted earthworms, which have simple one-dimensional body structure and move by propagating the contraction wave, as our model and proposed a decentralized control scheme in which TEGOTAE was fully exploited. The validity of the proposed control scheme was confirmed via simulation.

Norimasa Nomioka, Intelligent Systems and Informatics Lab, University of Tokyo
"Emergent locomotion patterns of soft-bodied robots with information maximization"
The ability to adaptively move in complex environment is a key skill for robots as well as animals. The maximization of predictive information is a prospective control method for nonlinear complex system. In this study, we designed information-based controller for soft-bodied robots. In both computer simulations and robot experiments, robots explored locomotor behaviors and could produce multiple periodic motion patterns. We analyzed emergent periodic patterns in terms of travel distance and mutual information between sensory input and motor output. The result suggests that large mutual information can lead to large travel distance.


12:00 - 13:15 | Lunch

13:20 - 14:50 | Session 3B

Rob Wood, School of Engineering and Applied Sciences, Harvard University
"Highly dynamic locomotion of an insect-size robot"
Through the use of a unique meso-scale manufacturing process based upon folding of quasi-2D composite materials, we have demonstrated high-speed (>10 body lengths per second) locomotion of a 2-gram legged robot. Robot complexity (e.g., measured by actuated degrees of freedom) typically decreases with reduced size. Our manufacturing methods, however, buck this trend and allow us to create fully-actuated robots and perform studies on gaits, gait transitions, and control. Furthermore, this robot serves as a platform to experiment with novel sensors, computation architectures, and power solutions that all must fit within the strict size, weight, and power limits of the insect-size device. This talk will discuss the development of this platform, the results of gait control studies, climbing and adhesion, integration of sensors, computation, and power, and steps towards full autonomy. 


Ron Fearing, Electric Engineering and Computer Science, University of California, Berkeley
Maneuverability in Under-actuated Dynamic Legged Locomotion"
Minimally-actuated palm-size robots are capable of running at speeds greater than 4.5 meters per second (45 body lengths per second), with leg stride rates of greater than 40 Hz.  However, maneuverability remains a challenge while running, as foot-ground interactions can work to stabilize body yaw, reducing turn rates.  Recently, we found that certain roll-oscillation modes can be used for continuous high-speed turning. Other continuous turning modes have also been identified, such as modulating foot contact location through foot compliance, and controlling differential leg velocity. For the small minimally-actuated robots examined, the dynamically enhanced roll-steer mode showed the best turning rate, of over 8 degrees per step, but only appears at certain running frequencies.  Inter-stride phase and velocity control appears promising as a mode for in-plane maneuverability for under-actuated robots.

Fumiya Iida, Engineering, Cambridge University
"Adaptation of locomotion behaviors in model-free robot evolution"
Although the conventional evolutionary robotics researches have been constrained in virtual environment, we are now able to investigate artificial evolution in the real world owing to the recent advances in robotic and fabrication technologies. We developed a robotic system that is capable of generating a large variety of locomotion robots autonomously, and visually analyzing the performance of them for the design improvement in an iterative fashion. The system is composed of a mother robotic arm equipped with a computer vision setup, and it handles and assembles sensory-motor modules into a variety of child locomotion robots. We have so far experimented with this system successfully generating over 500 individual child robots autonomously with a large morphological diversity, and demonstrated that model-free evolution can improve locomotion performance significantly. In this presentation, we report the challenges and perspectives of model-free evolutionary robotics based on our initial experiments.

Alex Sprowit , Physical Intelligence Department, Max Planck Institute
"Spring loaded leg design for dynamic locomotion: potential improvements in energy efficiency"
Cost of transport (COT) of legged animals has been documented for many species, and trends have been documented. If we compare legged machines to COT characteristics of animals, we find that morphological design in machines plays an important role. Not unlike animal taxonomic classification, currently different classes of robots emerge. I am focusing here on recorded walking data of ATRIAS robot, a bipedal machine with mechanical springs serially mounted to its actuators (SEA). The goal is to create an understanding of the effect of mechanism design, i.e. SEAs, on COT. Typically models are applied to make those estimations, with the disadvantage of inherent simplifications. Here I will show that recorded robot data in combination with motor models can give additional, important pointers.

Matteo Cianchetti (on behalf of Francesco Giorgio-Serchi), 
Cephalopod-inspired soft robots: design criteria and modelling frameworks "
In this talk we present our work on the development of aquatic vehicles inspired by cephalopods (i.e. octopuses and squids). These cephalopod-inspired, soft-bodied vehicles entail a hollow, elastic shell capable of performing a routine of recursive ingestion and expulsion of discrete slugs of fluids via the actual inflation and deflation of its own structure. The design development is complemented with an extensive modelling analysis aimed at aiding the process of mechanical optimization as well as providing an advanced tool for biomechanical studies of living cephalopods.

14:50 - 15:00 | Break / Refeshments

15:00 - 17:00 | Lab tours - Choice 1: MIT (Herr's LabHogan's LabKim's Lab), Choice 2: Harvard (Lauder's labWood's Lab

18:00 - 21:00 | Dinner Cruise - Boarding At: Rowe's Wharf, Gate B, Boston