Inside the Biorobotics Lab at Carnegie Mellon University, a wall of remembrance displays 15 failed, broken, burned out, and obsolete mechanical snakes. On the floor, a snake robot named Uncle Sam demonstrates the current state of the art. Prodded by the forward tilt of a joystick, Uncle Sam slithers ahead at a few miles per hour. Shift the stick to the right, and the robot’s 16 segments begin to sidewind like an elliptical, flexible corkscrew.
Upon reaching a 4-inch-diameter pipe protruding from the wall, the snake robot pauses, raises its flashlight head 8 inches off the ground, and begins to wriggle through. On the other side of the wall, discovering an 8-inch gap between the pipe exit and a table, it anchors its tail inside the pipe and stretches its body across the chasm. The table is too small for its 38-inch length, but Uncle Sam curls the rest of its body onto the platform while investigating a pole rising toward the ceiling. The robot then wraps itself around the pole, and, after a couple of false starts, begins to roll its entire body to the top of the 3-foot pillar.
Such maneuvers are certainly impressive. But like the 15 broken snakes in the display, Uncle Sam represents just another rung on the ‘bots’ evolutionary ladder, says lab director Howie Choset, an associate professor of robotics. From Tokyo to Philadelphia, a coterie of engineering professors is devoted to the continual improvement of snake robots, sometimes called “snakebots.” This determined group of researchers has seen success come slowly when dealing with the challenges of a form-changing device with no wheels to propel it and an infinite range of possible movements. But at the top of the pole, they see the prize: lifesaving missions in environments too confining and complex for ordinary robots, and too difficult or dangerous for humans to explore.
Last year marked an impressive milestone in the 30-year history of snake robots. A 1-centimeter-diameter snakebot invented in Choset’s lab and commercialized by Cardiorobotics Inc., a company he co-founded in 2005, made its first surgical explorations of a living heart, entering the patient’s body through a small incision. The heart patient was able to leave the hospital the next day, instead of enduring a weeks-long recovery from open-chest surgery.
Technically, snake robots are hyper-redundant devices because they move with so many internal degrees of freedom. To navigate to any point within reach, an arm — robotic or human — needs enough joints to provide six degrees of freedom. The human arm has a shoulder joint that can angle up and down (pitch), move left and right (yaw), and tilt side to side (roll) for three degrees of freedom. The elbow can pitch to add a fourth degree of freedom, and the wrist has the same three movements as the shoulder, to make a total of seven. With one more degree of freedom than the required six, the human arm is redundant. That extra degree of freedom is what allows a person to reach for a door handle with the elbow in a range of positions.
Uncle Sam’s joints alternate with pitch or yaw for a total of 16 degrees of freedom, making it hyper-redundant. A hyper-redundant robot that can move by changing its shape is a snake robot. Other hyper-redundant robots resemble an elephant’s trunk more than they do a snake; their base is anchored in place or moved about independently of the multijointed arm.
The history of snakebots began one day in 1971, when Shigeo Hirose, then a young mechanical engineer at the Tokyo Institute of Technology, walked into a unique Tokyo restaurant that served snakes and handed over about $15 in yen. Instead of walking out with an order of snake and vegetables, however, he left with a box of several wriggling snakes. Back at his laboratory, he used cameras, electrodes, and force sensors to analyze their movements. Among other observations, Hirose determined that snakes do not slither in a sine wave, as biologists had long assumed, but in a variation of a sinusoid curve he named the “serpenoid curve.” With three months of intense effort, he built the world’s first snake robot that could propel itself, powered only by the servomotors within every segment continually changing the angles of the joints.
Since then, Hirose, now the 63-year-old director of the institute’s Hirose-Fukushima Robotics Lab, has built more than 20 snake-robot prototypes. They range from a swimming snakebot to a hydraulic hyper-redundant robot so powerful that Nissan uses a variation of it on an assembly line to help lift heavy parts into the undercarriage of automobiles. Carnegie Mellon’s Choset says that the Japanese pioneer “has been hitting home runs each year for nearly 40 years.” And Hirose is not slowing down. “One or two years from now, I will show you much more interesting robots,” he promises.
In the United States, snake robot research was centered at the California Institute of Technology starting in the late 1980s and has since spread eastward. Choset’s Biorobotics Lab, part of the Robotics Institute at Carnegie Mellon in Pittsburgh, is the most active today. Several graduate students work with Choset in snakebot research. At the University of Pennsylvania, in Philadelphia, mechanical engineer Mark Yim and his students have developed modular robots that can be assembled — indeed can reassemble themselves — into snake shapes.
Today, Yim is working with Georgia Tech mechanical engineer/mathematician/biologist David Hu to make his snake robots function more like real snakes. Yim explains that early snakebot builders put passive wheels on their robots to allow them to slither. Freely turning wheels have the advantage of high sideways friction and low forward friction, but they require a smooth floor and prevent other snake gaits, such as sidewinding. “The problem is: Snakes don’t have wheels, so how do they move?” asks Yim. Hu’s research shows that the belly scales of a snake create extra resistance in the transverse direction — “weird, anisotropic friction,” in Yim’s words. One Penn mechanical-engineering graduate student is developing a mechanical snakeskin analogue that mimics this property. “I expect to have a major breakthrough soon,” says Yim.
Of course, a snake robot is not limited to the movements of its biological brethren. Both Hirose and Yim have taught snake robots to bite their own tails, so to speak, in order to roll in a loop. The result looks like a tank tread that has gone AWOL, moving about without its chassis. Yim reports that by maintaining the center of gravity in front of the point where the loop makes contact with the ground, the snakebot represents the fastest, most efficient mode of locomotion. “In effect the robot is continuously falling,” he explains.
The versatility, flexibility, and small diameter of snake robots make them ideally suited to inspection of pipes and ducts. Some of Choset’s snakebots can climb a vertical pipe from the outside or from within. His students and robots have made two trial runs working inside pipes and turbines at power plants. HiBot, a spinoff of the Hirose-Fukushima Robotics Lab, has begun selling a snake robot for the inspection of ductwork.
But the holy grail for snake robot engineers is search and rescue. Snakebots are physically capable of moving through, around, or over almost any obstacle that a natural or man-made disaster might present, such as collapsed buildings with confined spaces that human rescuers cannot reach. “Snakes are fantastic at going through a cluttered environment,” says Yim. Entirely sealed in a skin, a snake robot also can be made waterproof and dustproof. Uncle Sam is so durable that it has launched itself off a 10-foot drop onto a solid floor and slithered away. And since a snakebot is made up of many identical modules, “if one joint breaks, the whole thing keeps going,” says Choset. “We’ve had as much as half of a robot broken, and it still keeps going.”
Choset’s vision is that in the aftermath of future earthquakes, teams of snakebots will be “going everywhere, finding people and saying, ‘Help is on the way.’” Already the latest iteration of his snake robot has not only the usual light and camera on its head but also ports for microphones and speakers to allow communication with victims. “Search and rescue is Howie’s one true passion in robotics,” says Carnegie Mellon graduate student David Rollinson. “In his lifetime, he wants to have a snake crawl through a rubble pile and rescue someone.”
Clearly, snake robots capable of searching through rubble in an area contaminated with radiation would have been useful in helping Japan cope in March and April with the earthquake and tsunami, which left more than 25,000 dead and destroyed tens of thousands of homes. But Shigeo Hirose says the Japanese government never anticipated that a nuclear accident would be part of such a disaster, “so little attention was paid” to developing snakebots that could respond. Relatively little research funding has been available. Now, Hirose says in an email, “of course it will be changed completely and I hope funding for search and rescue will grow soon.”
Funding is not a problem for another snakebot market: medical devices. Carnegie Mellon spin-off Cardiorobotics has raised more than $11 million in venture funding and is about to enter trials for FDA approval of its hyper-redundant robotic catheter. With 102 joints, five of which can be controlled at the same time, the device significantly improves upon the maneuverability of existing laparoscopes and endoscopes. A typical rigid laparoscope needs to travel though the body in a straight line, while a flexible endoscope lacks precise control. “It’s like pushing a wet spaghetti noodle,” says Choset. The Cardiorobotics device, driven by a joystick, can enter a small incision in the abdomen and wend its way to map all sides of the heart or even perform surgical repairs.
The future of snake robots depends more on mathematics than money, however. To move independently through an unstructured space such as a rubble pile, a snake robot must map and choose among an infinite number of routes. Then it must negotiate the chosen route by using a gait selected from an infinite number of possible postures. “It takes tons of mathematics,” says Choset, who has a Ph.D. in mechanical engineering but has just published an article in an applied mathematics journal. Among the fields of study essential to programming snake robots, he lists differential geometry, topology, Lagrangian mechanics, probability, and statistics, particularly filtering and estimation.
Matthew Tesch, a Carnegie Mellon robotics graduate student, is using complicated mathematical formulas to answer the simplest of questions for snake robots: What is up and down? When a robot has 16 modules, each of which may be moving in a different direction at any given moment, that question is anything but simple. “It seems so first-grade-ish, but I had to go back to the beginning,” says Tesch. The systems-engineering graduate says that the effort is worth it because a snake robot will always seem little more than a slithering bundle of unrealized potential “until we can just put it in a rubble pile, press a button, and have the robot search for survivors on its own.” For now, the hope that snake robots will leap the gap between their current promise and ultimate utility rests almost entirely upon a group of committed academic engineers and their energetic students, who see nothing to fear in nature’s sinuous creatures.
Don Boroughs is a freelance writer based in South Africa.