Octopuses are some of the most complex, bizarre and intelligent creatures in the sea. They can squeeze through holes smaller than a quarter, pull with hundreds of pounds of force, change the color and texture of their skin in an instant and, with their walnut-sized brains, figure out how to open a childproof pill bottle to reach a tasty morsel of crab. With such an impressive array of skills, it was only a matter of time before engineers started asking: Could we make a robot that behaves like an octopus?
The OCTOPUS Integrating Project is one group that is trying to answer just this question. This multi-institution, international collaboration is working toward a fully autonomous robotic octopus that could, like a real mollusk, accomplish feats that no hard-jointed bot ever could.
Cecilia Laschi, a biorobotics researcher at the Sant'Anna School of Advanced Studies in Pisa, Italy, has been coordinating the effort. She and her colleagues completed a prototype disembodied octopus arm in 2010 and are now building the remainder of the body—from mantle top to arm tip. Their goal is to create a robot that will move like an octopus underwater and maneuver into tight spaces. It could be invaluable for search and rescue and exploration. But perhaps the most exciting aim of the project is to prove that creating an entirely soft-bodied robot is possible.
I visited Laschi and her colleagues at their laboratory in Livorno, a short train ride outside of Pisa. One of Laschi's researchers, biorobotics Ph.D. Laura Margheri, led me into the small, seaside facility. Inside, the many side doors were flung open, letting in a gentle Ligurian breeze and a view of boats in the harbor. At rows of workstations, some graduate students and postdocs labored away at their computers and tinkered with prototypes. In the center, at the head of the lab, was a large, well-appointed saltwater tank complete with rocks, starfish and one aging—but active—octopus. This mascot goes by the name of Andreino (an homage to a former colleague who caught him).
Margheri has been testing the octopus's natural abilities in hopes of mimicking them in the robots. She has rigged some clever experiments to see just how far octopuses can stretch their muscular hydrostat arms. In one exercise, an octopus uses a long arm to retrieve a treat from inside a long tube—something it can learn to do in just five training sessions over a couple of days. Margheri then places the food even farther down the tube and measures just how much these tonguelike arms can stretch. The arms, it turns out, can extend to about double their original length—an engineering challenge indeed.
The project leader's own background is in more traditional robotics. “I'm used to robots that have rigid links,” Laschi explained. After working with neuroscientists and learning more about our own brains and how we coordinate our bodies, she started to feel a little frustrated by the traditional rigidity of classic robots and the absence of structures like muscles. So she and some more bio-oriented collaborators started to plan a daring soft-bodied robot project. And what better model than the octopus? “All biological systems have some soft material,” she said, “but the octopus is very special because it has only soft material” (except for the beak, of course). “So we took it as the extreme—if you study this end of the spectrum, you can solve the others.” A group at Harvard University has developed a four-legged octopus-inspired robot that can inch along on land and even (slowly) color camouflage. But it still requires tethers to air and liquid pumps and external controls to guide it.
To mimic the octopus more completely—from the inside out—Laschi's team used ultrasound technology to get a rare internal view of the octopus's arm and all its muscles at work. This perspective helped to clarify “the secret of the movement of the octopus arm,” chimed in Matteo Cianchetti, another bioroboticist in Laschi's lab. In the absence of a skeleton, three groups of muscles give the octopus arm its flexibility and structure, allowing it to change direction, length and even stiffness. To replicate the muscles, the researchers are using cables and springs forged from shape-memory alloys that bend when heated by an electric current and subsequently return to their original shape.
Although the muscles themselves can extend far beyond their typical length, the central nerve cord in each arm cannot. Instead each nerve bundle is folded in a zigzag configuration, somewhat like an accordion, so it can unfold as the arm extends. Taking a cue from living octopuses, Margheri and her team are packing processing wires into the center of the arm in a wavelike pattern.
Cianchetti showed me one of the silicone-skinned prototypes, which has a ghostly gray hue. By pulling on a few wires, he made it curl up into a spiral. I stuck my finger out, and the disembodied arm firmly wrapped its rubbery skin around my own with a disconcerting ease. By virtue of the arm's shape, proportions and the “musculature” inside, it naturally winds around whatever it is grasping. “It automatically adapts to that shape,” Cianchetti said. Great, I thought with a shiver.
The researchers are studding the robot arm with sensors that integrate tactile perception and hope to also add some kind of suckerlike appendages. These might not, however, behave exactly like a living octopus's suckers, which are strong but also versatile—able to rotate, fold and even taste the environment around them. Other groups, including Frank Grasso's lab at Brooklyn College, are developing more refined robotic suction cups. And the U.S. Army Research Laboratory, in collaboration with other scientists, is already 3-D printing superstrong, individually activated suckers.
Scientists working on robot octopuses choose their materials carefully, so the robots can perform well underwater for long periods without corroding. The silicone that the OCTOPUS Integrating Project team is using has almost the same density as water, so it is buoyant—just like a real octopus.
Because an underwater robot, no matter how impressive, is only really functional if it can get around, some of the lab members are investigating various forms of locomotion. Marcello Calisti, another bioroboticist, is tackling the walking problem. Most real octopuses walk more with their back arms while feeling around with their front ones. But for the artificial version, the roboticists might instead have it reach out with its front arms, attach some suckers, then pull, a strategy that will also help with exploration and determining directional movement.
Calisti's workstation was next to a half-full inflatable kiddie pool used to test underwater crawling and other tasks. Calisti showed me his current prototype, made out of hard-material motors and fixed cables, which had only six arms and still looked rather primitive. But there was something eerie about it when Calisti played back a video of it in action crawling along, spiderlike. “It's a little bit creepy,” he admitted. So far they have been able to program it externally, turn it loose in the pool, and watch it locate and retrieve objects. But the goal is to eventually put the command center inside of it (and give it the full eight legs and a totally soft body).
Crawling, of course, is not the octopus's only means of getting around. In the real world, octopuses opt for jet-powered swimming when they need to make a quick escape. At the other end of the lab, Francesco Giorgio Serchi was trying to re-create the octopus's propulsion system in silicone. The water jet that octopuses can propel themselves with creates a swirl of water known as a vortex ring. The real animal generates this force by using its mantle muscles to suck up and squirt out water through its funnel.
Scientists are only now figuring out the fluid dynamics of the vortex ring, which squid and a few other underwater animals also employ. The goal is to mimic this feat of biophysics to someday propel small submarines or autonomous vehicles, Giorgio Serchi noted. As he pointed out, adapting this aquatic locomotion for our purposes would be a big step forward. With current technology, most “every kind of propulsion in the water environment is continuous,” he said. Propellers and even water jet boats generate a constant motion. In contrast, octopus-inspired propulsion “would be the first example where you actually use a discontinuous jet.” And not just for novelty's sake. “It's interesting,” he remarked, “because it appears that it is especially efficient.” Re-creating a vortex ring could give underwater vehicles extremely efficient acceleration.
But you can't just tie on a big turkey baster, fill it with water and squeeze. The octopus's system is a bit more nuanced and finely tuned than that. “Certainly the most complex aspect here is reproducing this capability of his to just contract a little bit, the width of the mantle, and change significantly the volume inside, which gets displaced,” Giorgio Serchi said. “It's a challenge.”
The octopus, however, is doing this with ease. So Giorgio Serchi decided not to reinvent the proverbial wheel. Instead he took a cast of a real octopus mantle and then reconstructed it in silicone. He showed me the detailed model. There were even cavities where the organs go, which, for now, he had filled with electronic components. “It's a big approximation,” he conceded. But the results should also help inform biologists about how these cephalopods swim.
The next step for the roboticists is adding flexible intelligence to their creation. Aside from the engineering challenges, Laschi and the rest of the OCTOPUS Integrating Project researchers are vexed with a biological question, “How can an animal with a relatively small brain control such a huge amount of physical freedom and sensory data?” The jury is still out on how the animals do it, but that is not going to stop the engineers. So Laschi has two words: embodied intelligence. This means that each part of the body—octopus or robot—is, at least in part, in control of itself.
To run all those arms so exquisitely, “there must be a lot of embodied intelligence,” she said. “Each arm has many neurons and controls a good part of the movements, but we don't have a real model from neuroscience of how it works.
Not only does neuroscience fail to explain these abilities, but traditional robotics also comes up short. Robotic control has been based on rigid, finite movements. But what do you do when you have a near infinite range of motion with multiple parts? This is, of course, precisely the dilemma that biologists have been coming up against as well when looking at the octopus itself.
In searching for solutions, Laschi and her team have turned to a common evolutionary answer: learning. Just as we and many animals—including cephalopods—learn at a young age how to control our limbs, so, too, will these soft-bodied robots. This approach is appealing, in part, because it does not require exhaustive modeling. Over time the robot octopus could learn to apply a single movement to many different tasks and to combine various movements for more complex challenges. If it encounters an obstacle—a rock on the seafloor, for example—it might run through a variety and combination of different known commands. Once it finds the movement or combination of movements that allow it to surmount the rock, it will remember to use the same techniques when faced with a similar roadblock. In this sense, it should learn somewhat as we do and eventually become more “intelligent.”
To create an intelligent robot, the team first needs to engineer body feedback systems, including adding more sensors in the arms to detect how much the limbs have extended or contracted. The scientists might be able to use the shape-memory alloy spring itself as a sensor. “We will have both tactile sensors and some kind of position sensors,” Laschi said.
Engineers have a long way to go in their quest to transform the amazing octopus into a robot. In the years to come, their achievements and failures alike will provide new insights into the biology of one of the ocean's craftiest creatures, as well as help robotics surpass the limitations of rigid structures to embrace smarter and far more flexible forms.
