Biologist Robert Full shares his fascination with spiny cockroach legs that allow them to scuttle at full speed across loose mesh and gecko feet that have billions of nano-bristles to run straight up walls. His talk, complete with wonderful slow-mo video of cockroach, crab and gecko gaits, explains his goal of creating the perfect robotic “distributed foot.” (Recorded February 2005 in Monterey, California. Duration: 19:24.)
Watch Robert Full’s talk on TED.com, where you can download it, rate it, comment on it and find other talks and performances.
Read more about Robert Full on TED.com.
Robert Full — Inspired by Nature
I want you to imagine that you’re a student in my lab. What I want you to do is to create a biologically inspired design.
(slide: “Inspired by Nature
Biologically Inspired Design
Fully 3D Dynamic Parameterized Contact Model
Professor Robert J. Full
University of California, Berkeley
Department of Integrative Biology
And so here’s the challenge. I want you to help me create a fully 3-D, dynamic, parameterized contact model.The translation of that is “Could you help me build a foot?” And it is a true challenge, and I do want you to help me. Of course, in the challenge there is a prize. It’s not quite the TED Prize, but it is exclusive: a T-shirt from our lab. So please send me your ideas about how to design a foot.
Now if we want to design a foot, what do we have to do? We have to first know what a foot is. If we go to the dictionary, it says:
(slide: “Foot noun pl. feet — the lower extremity of the leg that is in direct contact with the ground in standing or walking.”)
It’s the lower extremity of a leg that is in direct contact with the ground in standing or walking. That’s the traditional definition. But if you wanted to really do research, what do you have to do? You have to go to the literature and look up what’s known about feet. So you go to the literature.
(slide: page from Dr Seuss, “Left foot, right foot, foot foot foot, How many, many feet you meet.”)
(laughter) Maybe you’re familiar with this literature. The problem is, there are many many feet. How do you do this? You need to survey all feet and extract principles of how they work. And I want you to help me do that in this next clip. As you see this clip, look for principles, and also think about experiments that you might design in order to understand how a foot works.
(clip: shots of many different kinds of feet in rapid succession, including flippers, insects, prostheses, arachnids, sea creatures, mountain climbers, etc.
“Pauline Jennings, Music by Sean Clute” — then credits)
See any common theme? Principles? What would you do? What experiments would you run?
Our research on the biomechanics of animal locomotion has allowed us to make a blueprint for a foot.
(slide: “Blueprint for a Foot”; “Foot Design Inspired by Nature
Prototype 1.0 Feb. 2005″ — rendering of an insect-like foot with graph at bottom labelled “rough — smooth”)
It’s a design inspired by nature, but it’s not a copy of any specific foot you just looked at, but it’s a synthesis of the secrets of many, many feet.
(slide: “Go Anywhere!” pictures of lizard, spider, other animals on left labeled “Animals”, pictures of surfaces and enviroments on right labeled “Substrates — Bark, sand, rocks, dirt, leaves”
“Animals locomote on substrates that vary in:
1. the probability of surface contact,
2. movement and
3. the type of footholds present”)
Now it turns out that animals can go anywhere. They can locomote on substrates that vary as you saw. In the probability of contact, the movement of that surface, and the type of footholds that are present. If you want to study how a foot works, we’re going to have to simulate those surfaces, or simulate that debris. When we did that, here’s a new experiment that we did-
(slide: looped film of running spider — “Grass Spiders — Distributed Foot
99% of Contact Removed!
70 body lengths/ sec on Mesh!
Human Equivalent of Running at 300 mph
Spagna, Goldman, and Full, 2005″)
We put an animal and had it run, this grass spider, on a surface with 99% of the contact area removed. But it didn’t even slow down the animal. It’s still running at the human equivalent of 300 miles per hour. Now how could it do that?
(Runs in slow motion — “Grass Spiders — Distributed Foot
Leg Acts as a Foot!
1/50 of Actual Speed
Spagna, Goldman, and Full, 2005″)
Well, look more carefully. When we slow it down 50 times we see how the leg is hitting that simulated debris. The leg is acting as a foot. And in fact,
(left side of slide switches to spraph showing frequency of parts of the leg hitting the surface)
The animal contacts other parts of its leg more frequently than the traditionally defined foot. The foot is distributed along the whole leg.
You can do another experiment where you can take a cockroach with a foot:
(slide: running cockroach: “Remove Traditional Feet — No Difference
Runs at High Speed Even Without ‘Feet’
Spagna, Goldman, and Full, 2005″ — shows cockroach with feet, and feet removed — running cockroach is without feet)
And you can remove its foot. I’m passing some cockroaches around. Take a look at their feet. Without a foot, here’s what it does: It doesn’t even slow down! It can run the same speed even without even that segment, no problem for the cockroach, they can grow them back, if you care. How do they do it? Look carefully, this is slowed down 100 times:
(slide — slomo cockroach — same text, film is “Slowed 100x”)
And watch what it’s doing with the rest of its leg. It’s acting, again, as a distributed foot. Very effective.
Now, the question we had is — How general is a distributed foot? And the next behavior I’ll show you of this animal just stunned us the first time we saw it.
(film — “Bipedal Octopus Disguised as a Rolling Coconut
Christina Huffard & Sea Studios, Inc., Monterey (cameraman Bob Cranston)”, film shows running octopus)
Journalists, this is off the record, it’s embargoed — Take a look at what that is! That’s a bipedal octopus disguised as a rolling coconut. It was discovered by Christina Huffard and filmed by Sea Studios, right here from Monterey. We’ve also described another species of bipedal octopus:
(film: “Bipedal Octopus Disguised as Floating Algae
Christina Huffard and R.J. Full, 2005″ — another running octopus)
This one disguises itself as floating algae. It walks on two legs and it holds the other arms up in the air so that it can’t be seen. (applause) And look what it does with its foot to get over challenging terrain.
(“Bipedal Octopus Using Distributed Feet
C. Huffard and R.J. Full, 2005″ — octopus running over coral, etc.)
It uses that beautiful distributed foot to make it as if those obstacles are not even there. Truly extraordinary.
In 1951, Escher made this drawing. He thought he created an animal fantasy.
(slide — “Art Imitates Life, Escher 1951 House of Stairs”)
But we know that art imitates life, and it turns out nature, 3 million years ago, evolved the next animal, it’s a shrimp-like animal called the stomatopod, and here’s how it moves on the beaches of Panama:
(film — “The Ultimate Distributed Foot”, (credits illegible), shrimp thing rolling across surface with entire body)
It actually rolls, and it can even roll uphill. It’s the ultimate distributed foot, its whole body in this case is acting like its foot.
So, if we want to, then, to our blueprint add the first imporant feature, we want to add distributed foot contact.
(back to blueprint slide — “Distribute Foot Contact”)
Not just with the traditional foot, but also the leg, and even of the body.
Can this help us inspire the design of novel robots? We biologically inspired this robot, named RHex, built by these extraordinary engineers
(slide — picture of “RHex”, “Legged Robot — Biologically Inspired Robot
Martin Buehler — Boston Dynamics
Dan Koditschek — University of Pennsylvania
Al Rizzi — CMU
No sensing of environment”)
over the last few years. RHex’s foot started off to be quite simple, then it got tuned over time, and ultimately resulted in this half circle.
(slide — photos of the foot as it evolved — “Evolution of a Distributed Foot — Martin Buehler, Boston Dynamics”)
Why is that? The video will show you. Watch where the robot, now, contacts its leg in order to deal with this very difficult terrain:
(video — “Distributed Foot in Action” — Robot running over rocky terrain)
What you’ll see, in fact, is that it’s using that half circle leg as a distributed foot. Watchi it go over this.
(robot climbs over logs)
You can see it here well on this debris. Extraordinary. No sensing, all the control is built right in to the tuned legs. Really simple, but beautiful.
Now, you might have noticed something else about the animals when they were running over the rough terrain. And my assistant’s going to help me here. (approaches boy at side of stage) When you touch the cockroach leg — (to Chris Anderson) — Can you get the microphone for him? When you touch the cockroach leg, what do they feel like? Did you notice something?
It’s spiny, right? It’s really spiny, isn’t it? It sort of hurts. Maybe we could give it to our curator (hands cockroach leg to Chris A.) and see if he’d be brave enough to touch the cockroach. (laughter)
(Chris A.) “Did you touch it?”
So if you look carefully at this, what you see is that they have spines
(slide — “Tuned Spines
Spines Increase Effectiveness of Distributed Foot”
photo of cockroach leg)
and until a few weeks ago, no one knew what they did. They assumed that they were for protection and for sensory structures.
(slide — “Tuned Spines — Leg Segment of American Cockroach
Easily collapse in one direction to pull leg out of debris, but are stiff in the other to provide a foothold
Spagna, Goldman, and Full, 2005″)
We found that they’re for something else — here’s a segment of that spine. They’re tuned such that they easily collapse in one direction to pull the leg out from debris, but they’re stiff in the other direction so they capture the disparities in the surface.
Now crabs don’t miss footholds because they normally move on sand, until they come to our lab.
(slide & film loop — “Ghost Crab — No Spines
Running Performance on Mesh Decreased
0.15 m/sec on mesh
4.00 m/sec on beach
Goldman, Basho, Spagna, and Full, 2005″)
And where they have a problem with this kind of mesh, because they don’t have spines. The crabs are missing spines, so they have a problem in this kind of rough terrain. But of course, we can deal with that, because we can produce artificial spines.
(slide — “Crabs — Add Artificial Spines
Tune Spines to Catch Simulated Debris, but to Collapse on Removal
Goldman, Lara, & Full, 2005″)
We can make spines that catch on simulated debris and collapse on removal to easily put them out. We did that by putting these artificial spines on crabs, as you see here, and then we tested them. Do we really understand that principle of tuning — the answer is yes!
(film loop of crab running more quickly across mesh-
“Ghost Crab — Artificial Spines
Running Performance on Mesh Increased
Goldman, Basho, Spagna, and Full, 2005
This is slowed down 20 fold, and the crab just zooms across that simulated debris. (laughter and applause) A little better than nature.
So, to our blueprint, we need to add tuned spines.
(slide — “Distribute Foot Contact — Add Tuned Spines”)
Now will this help us think about the design of more effective climbing robots? Well, here’s RHex-
(film of RHex trying to climb over rails and failing
“Robot Fails on Rails
RHex slips on smooth rails
Pei-Chun Lin, Daniel Koditschek, U. Penn”)
RHex has trouble on rails — on smooth rails, as you see here. So why not add a spine? My colleagues did this at U. Penn.
(“Robot — Add Spines
Increase Effectiveness of Climbing?
Pei-Chun Lin, Daniel E. Koditschek, U. Penn”)
Dan Koditschek put some steel nails — very simple version — on the robot — and here’s RHex, now, going over those steel — those rails. No problem! How does it do it? Let’s slow it down and you can see the spines in action. Watch the leg come around,
(slide — “Robot with Spines
Spines catch to give better foothold
Pei-Chun Lin, Daniel Koditschek, U. Penn”
slomo video of robot climbing)
and you’ll see it grab on right there. It couldn’t do that before, it would just slip and get stuck and tip over. And watch again, right there — successful.
Now just because we have a distributed foot and spines doesn’t mean you can climb vertical surfaces.
(video — “RHex Slips during Climbing”, RHex sliding down inclined plane)
This is really, really difficult. But look at this animal do it –
(slide — “Feet Interacting with Surface, Cockroach Climbing a Smooth Vertical Metal Plate
Successful Rapid Climbing on Smooth Surface
Spagna, Goldman, and Full 2005″ — shows cockroach climbing)
One of the ones I’m passing around is climbing up this vertical surface that’s a smooth metal plate. It’s extraordinary how fast it can do it — but if you slow it down, you see something that’s quite extraordinary. It’s a secret-
(“Feet interacting with Surface
Cockroach Climbing a Smooth Vertical Metal Plate
Effective climbing achieved by slipping or “Swimming”
Surface behaves as Fluid
Distributed Foot like a paddle
Spagna, Goldman, & Full 2005″)
The animal effectively climbs by slipping and look — and doing, actually, terribly, with respect to grabbing on the surface. It looks, in fact, like it’s swimming up the surface. We can actually model that behavior better as a fluid, if you look at it. The distributed foot, actually, is working more like a paddle.
(slide — lizard running — “Surfaces as Fluids — Lizard Running Rapidly on Sand
Feet Behave as Paddles Even on Land
Goldman, Koeff (?), and Full, 2005)
The same is true when we looked at this lizard running on fluidized sand. Watch its feet. It’s actually functioning as a paddle even though it’s interacting with a surface that we normally think of as a solid. This is not different from what my former undergraduate discovered when she figured out how lizards can run on water itself.
(slide — lizard running on water in slomo — “Surfaces as Fluids
Lizard Running Rapidly on Water!
Harsh & Lauder”)
Can you use this to make a better robot? Martin Buehler (?) did, who’s now at Boston Dynamics-
(video of robot paddling through the water)
He took this idea and made RHex to be Aqua RHex. So here’s RHex with paddles, now converted into an incredibly maneuverable swimming robot.
For rough surfaces, though, animals add claws. And you probably feel them if you grabbed it. (to Chris A.) Did you touch it?
(C.A) — I did.
(slide — Cockroach climbing — “Claws for Climbing Rough Surfaces
Cockroach running on vertical surface — balsawood (Slowed 20x)”)
And they do really well at grabbing onto surfaces with these claws. Mark Cutkosky at Stanford University, one of my collaborators,
(slide — “New Way to Make Feet!
Mark Cutkosky — Stanford University
Shape Deposition Manufacturing
allows variable stiffness materials and simple embedding”)
is an extraordinary engineer, who developed this technique called Shape Deposition Manufacturing, where he can imbed claws right into an artificial foot. And here’s the simple version of a foot for a new robot that I’ll show you in a bit.
So, to our blueprint, let’s attach claws.
(slide — blueprint again — “Distribute Foot Contact
Add Tuned Spines
Now if we look at animals, though, to be really manuverable in all surfaces, the animals use hybrid mechanisms-
(slide — “Hybrid ‘Feet’
To maneuver on all surfaces animals use hybrid mechanisms including claws, spines, hairs, pads, glues and capillary adhesion” — pictures of different feet and an ant)
-that include claws and spines and hairs and pads and glue and capillary adhesion and a whole bunch of other things. These (refers to pictures of feet) are all from different insects. There’s an ant crawling up a vertical surface. Let’s look at that ant.
(slide — close up of foot)
This is the foot of an ant. You see the hairs and the claws and this thing here. This is when its foot is in the air. Watch what happens when the foot goes onto your sandwich.
(slide — second view of ant foot, compared with the first)
You see what happens? That pad comes out. And that’s where the glue is. Here from underneath is an ant foot-
(slide — “Pad Extension Couled with Claw Movement
Ant Foot on Superstring
Federle (?) & Full” — shot of ant foot on clear surface from underneath)
-and when the claws don’t dig in, that pad automatically comes out without the ant doing anything. It just extrudes. And this was a hard shot to get — I think this is the shot of the ant foot on the superstrings. So it’s pretty tough to do. This is what it looks like close up — here’s the ant foot — and there’s the glue.
(slide — “New Two Phase Glue
Glue A Two Phase Mixture
Federle & Full”
-two shots of ant foot, and glue that comes out of pad)
And we discovered this glue may be an interesting two — phase mixture. It certainly helps it to hold on.
So — to our blueprint we stick on some sticky pads.
(“Distribute Foot Contact-
Add Tuned Spines
Stick on Sticky Pads”)
Now you might think for smooth surfaces we get inspiration here-
(slide of Spiderman climbing a wall)
Now we have something better here –
(shot of gecko — “Nanotechnology in Nature”)
The gecko’s a really great example of nanotechology in nature. These are its feet –
(shot of lots of gecko feet from below — with the words “Hairy Toes” popping in)
They’re — almost look alien. And the secret, which they stick on with, involves their hairy toes. They can run up a surface at a meter per second, take 30 steps in that one second-
(slide — “Gecko Feats
Rapid Climbing on Smooth Surfaces
30 steps per second
Attaches in 8 msec
Detaches in 16 msec”
– gecko climbing)
You can hardly see them. If we slow it down, they attach their feet at 8 milliseconds, and detach them in 16 milliseconds. And when you watch how they detach it, it is bizarre.
(slo-mo closeup of gecko foot)
They peel away from the surface like you’d peel away a piece of tape. Very strange. How do they stick?
(“Nano-sized Split Ends
Tour of a Gecko Foot
1 million foot hairs (setae) — Single Sets — 1000 Spatular tips
Autumn et al, 2000, 2002″
-close ups of various parts of a gecko foot, captioned as above)
If you look at their feet, they have leaf-like structures called linalae with millions of hairs. And each hair has the worst case of split ends possible. It has a hundred to a thousand split ends, and that’s the secret. ‘Cause it allows intimate contact. The gecko has a billion of these 200 nanometer size split ends. And they don’t stick by glue, or they don’t work like velcro, or they don’t with suction. We discovered they work by intermolecular forces alone.
So, to our blueprint, we split some hairs.
(blueprint pops up again, this time with phrase “Split some hairs” added to bottom)
This has inspired the design of the first self-cleaning dry adhesive, the patent issued, we’re happy to say, and here’s the simplest version in nature, and here’s my collaborator Ron Fearing’s attempt at an artificial version of this dry adhesive made from polyurethane. And here’s the first attempt to have it work on some load.
(slide-”Biological Inspiration” — electron micrograph of “First Self-Cleaning, Dry Adhesive
Natural Lizard — Full (UC Berkeley), Autumn (Lewis & Clark)”
2nd photo is of weight hanging from adhesive)
There’s enormous interest in this in a variety of different fields, you could think of a thousand possible uses, I’m sure. Lots of people have, and we’re excited about realizing this as a product. We have imagined products, for example, this one,
(joke band-aid ad — with photo of exp. prototype — “Future Bio-Inspired Band-Aid? Johnson & Johnson
Now with Gecko Peel
Rows of Actual Gecko Hairs from Molting Animal for Demonstration
From Autumn, Gassett, and Schwab”)
We imagined a bio-inspired Band-Aid, where we took the glue off the Band-Aid, we took some hairs from a molting gecko, put 3 rolls of them on here, and then made this Band-Aid.
(film of hand beside picture of gecko foot, having a band-aid put on it)
This is an undergraduate volunteer — we have 30,000 undergraduates so we can choose among them — that’s actually a red pen mark. But it makes an incredible Band-Aid. So — it’s aerated, it can be peeled off easily, it doesn’t cause any irritation, it works underwater. I think this is extraordinary example of how curiosity based research — we just wondered how they climbed up something — can lead to things that you could never imagine. It’s just an example of why we need to support curiosity based research.
(film — band aid being removed)
Here you are, pulling off the Band-Aid.
So. We’ve redefined, now, what a foot is. The question is, can we use these secrets, then, to inspire the design of a better foot, better than one we see in nature? Here’s the new project — we’re trying to create the first climbing search and rescue robot — no suction or magnets-
(slide — “The First Climbing Search & Rescue Robot
NEW ROBOT! RiSE — No Suction or Magnets
Robot in Scansorial Environments
(…followed by research credits))
-That can only move on limited kinds of surfaces. I call the new robot RiSE, for Robot in Scansorial Environment — that’s a climbing environment — and we have an extraordinary team of biologists and engineers creating this robot. And here is RiSE.
(photos of robot added to slide — cut to film of robot climbing a surface — “RiSE’s First Steps Up Wall!”)
It’s six legged, and has a tail — Here it is on a fence and a tree. And here are RiSE’s first steps on an incline. You have the audio? …You can hear it go up. And here it is coming up at you, in its first steps up a wall. Now it’s only using its simplest feet here, so this is very new, but we think we got the dynamics right of the robot.
Mark Cutkosky, though, is taking it a step further. He’s the one able to build this shape deposition manufactured feet and toes — the next step is to make compliant toes-
(slide — “The Stanford Foot-bot!” “Next Step Compliant Toes with Spines, Claws, and set for Dry Adhesives” — diagram of robot feet — “soft flexures” — “nonlinear tip flexures buckle in compression” — “FEM model for toe deflection analysis”)
-and try to add spines and claws and set it for dry adhesives. So the idea is to first get the toes and a foot right, attempt to make that climb, and ultimately put it on the robot. And that’s exactly what he’s done — he’s built, in fact, a climbing foot-bot inspired by nature.
(“Professor Mark Cutkosky
Collaboration with RiSE Team
Design Inspired by Nature”)
And here’s Cutkosky and his amazing students’ design.
(film of robot crawling up a wall — comparative film to cockroach climbing up wall)
So these are tuned toes, there are six of them, and they use the principles that I just talked about collectively for the blueprint. So this is not using any suction, any glue, and it will ultimately, when it’s attached to the robot — it’s as biologically inspired as the animal — hopefully be able to climb any kind of a surface. Here you see it, next, going up the side of a building at Stanford.
(shot of robot foot going up brick wall, close up of feet attaching to brick)
It’s sped up, again, it’s a foot climbing, it’s not the whole robot yet, we’re working on it — now you can see how it’s attaching. These tuned structures allow the spines, friction pads, and ultimately the adhesive hairs to grab on to very challenging, difficult surfaces. And so they were able to get this thing — this is now sped up 20 times — can you imagine it trying to go up and rescue somebody at that upper floor?
(robot foot climbing wall to window)
OK? You can visualize this now, it’s not impossible — it’s a very challenging task. But more to come later.
(slide — “Design Secrets from Nature
1. Distribute Control to Smart Parts — Not all in Brain, but in Tuned Feet, Legs, and Body
2. Use Hybrid Solutions — Integrated & Robust
3. Do NOT Mimic Nature — Be INSPIRED by BIOLOGY and use these novel principles with the best engineering solutions to make something better than nature”)
We’ve gotten design secrets from nature by looking at how feet are built. We’ve learned we should distribute control to smart parts. Don’t put it all in the brain, but put some of the control in tuned feet, legs, and even body. That nature uses hybrid solutions, not a single solution, to these problems, and they’re integrated and beautifully robust. And third, we believe strongly that we do not want to mimic nature, but instead be inspired by biology, and use these novel principles with the best engineering solutions that are out there to make — potentially — something better than nature.
So there’s a clear message — whether you care about a fundamental, basic research of really interesting, bizarre, wonderful animals, or you want to build a search and rescue robot,
(slide — “Clear Message
Rescue Research Recovery
Must Preserve Nature’s Designs
Secrets Lost Forever”)
that can help you in an earthquake, or to save someone in a fire, or you care about medicine — we must preserve nature’s designs, otherwise these secrets will be lost forever. Thank you.
Transcription by Robert Thomas Carter