Hop . . . Hop . . . Hopbots!

Designers of small, mobile robots take cues from grasshoppers and frogs

Rush Robinett got the idea for his lab’s new robot while out on his father’s New Mexico farm. He was catching grasshoppers for trout bait and noticed that when he reached for the grasshoppers, they seemed to spring in random directions. They fell on their sides as they landed and then struggled back onto their feet before springing again.

In multiple-exposure photograph, robot prototype created by Sandia National Laboratories leaps over the head of Gary J. Fischer, one of its designers. Inset: The device on the ground. Randy Montoya/Sandia; inset: Gaye Garrison/Sandia

A fallen frogbot pushes with plastic levers to right itself after a hop. JPL-NASA/Caltech

A robot designer at Sandia National Laboratories in Albuquerque, Robinett realized that this kind of grasshopper mobility might be just the thing for developing new kinds of small, mobile robots. Program managers at the Defense Advanced Research Projects Agency (DARPA), in Arlington Va., had been interested in creating wide-ranging minirobots for years.

As both Robinett and his military sponsors knew only too well, small robots have a big problem.

Consider Sojourner, NASA’s celebrated hassock-size rover that poked around a tiny patch of Mars as part of the 1997 Pathfinder mission. According to the dictionary, to sojourn implies to stop and stay in one place for a spell. It seems that the small, wheeled robot did just that nearly every time a rock as big as itself blocked its way. Sometimes, it got hung up for days on end.

Not exactly the kind of get-up-and-go you want from a robot sent millions of kilometers to explore exotic worlds.

Agencies like NASA and the Department of Defense have been championing the idea of little robots because they can be cheap and portable but can still carry a payload-cameras, scientific instruments, or perhaps bombs. However, “as you get smaller, obstacles the size of yourself come along much more frequently,” explains Sandia’s Barry Spletzer. “If you’re an Abrams tank, there are not many obstacles your size, but if you’re an ant, every blade of grass is bigger than you are.”

While Sojourner was getting stuck on Martian rocks, two independent groups-the Sandia researchers and a joint team from NASA’s Jet Propulsion Laboratory (JPL) and the California Institute of Technology (Caltech), both in Pasadena-independently hit upon the same new concept for dealing with the problem of obstacles.

From the robot’s perspective, it goes something like this: Throw yourself as high and as far as you can in the general direction you want to go. After the crash landing, right and reorient yourself. Then, do it again until you reach your goal. In the past few months, each group has unveiled a prototype of what it intends to develop into a new generation of hopping robots.

The Sandia robot looks nothing like a grasshopper, but it sure can jump. “They’ve put an emphasis on really good thrusting. It’s amazing,” comments Joel Burdick, one of the designers of the rival JPL/Caltech machine.

Dubbed by its inventors as “the knight” -after the chess piece that leaps over others-the robot is basically a piston-driving combustion chamber mounted inside a spherical, plastic shell about the size of a grapefruit. To propel each of the half-kilogram robot’s leaps, a small charge of liquid propane or other fuel ignites within the chamber. This slams the piston against the ground. Typically, that push-off hurls the device about 1 meter into the air and a couple of meters away from its starting point.

The New Mexico team is also working on a powerful, 2.5-kg version of the robot, which has leaped 4 m high and 5 m away in preliminary tests, says Sandia engineer Gary J. Fischer.

When the hopper lands, the imbalance between its relatively lightweight, spherical top and relatively heavy, cone-shaped base causes it to automatically right itself. Then, a small, battery-powered motor rotates an internal, off-center weight to pivot the device around and realign it according to its built-in compass.

Like a tacking sailboat, the robot can head more-or-less in a chosen compass direction. “Even though individual hops are fairly inaccurate, over the long term you get where you’re going,” Spletzer says. The Sandia team figured that more sophisticated navigation would be unnecessary for the uses their sponsors had in mind for the robot and would only add to the complexity, weight, and cost.

Initially, DARPA officials envisioned a small, rugged, camera-equipped robot that soldiers or a SWAT team could toss into a building to snoop around. Then, another DARPA program took over funding for the robot’s development. This program’s goal is to create mobile antitank mines that can shuffle around the minefield and close up any gaps that enemy mine-clearing actions might have opened.

The energy-efficient knight can hop thousands of times on a 20-gram fuel tank, enabling it to wander for at least several kilometers before running out of gas. With that kind of range, the device might serve other applications as well. “My favorite is for planetary exploration,” Spletzer says.

The robot designers at JPL and Caltech found their inspiration in another hopper in the animal kingdom. They call their electrically powered invention “frogbot.”

By appearances alone, the 40-centimeter-tall aluminum and plastic gizmo seems more like the skeleton of some big-beaked bird than a frog.

Watch it move, however, and the reason for the frog moniker jumps out: A pair of hinged shafts, bent like a pair of squatting frog’s legs and tethered together at the knees by a strong spring, suddenly straighten. The action resembles a frog’s leap, except the legs share a single foot.

Since it falls over when it lands, frogbot is equipped with plastic levers for righting itself. The levers push against the ground, forcing the 1.5-kg device erect. Then the motor pivots the device on its base to point in the right direction and rebends the legs for the next leap.

In the prototype, an operator controls that direction, but future versions will contain an orientation sensor. “Mars, for example, has no [planetwide] magnetic field, [so] it is necessary to use a sensor that reads the position of the sun in the sky,” says JPL engineer Paolo Fiorini, who heads the project.

On Earth, frogbot lunges about 2 m per leap-roughly as far as the Sandia robot. But in the low gravity of Mars, it would go three times as far in a single bound.

So far, the frogbot prototype hops only a half-dozen times before it must recharge its batteries, says Fiorini. However, he notes, models ultimately sent to space will rely on solar power, which will enable them to roam indefinitely.

In 5 to 10 years, NASA could be sending hoppers into space by the hundreds, predicts Neville I. Marzwell, manager of JPL’s advanced concepts and technology innovation office. His office and the National Science Foundation are funding the JPL-Caltech hopper’s development.

Spacecraft would drop groups of hoppers onto planets, asteroids, or other bodies. The robots would then fan out to explore the surface. Compared with wheeled rovers, hoppers would have a better chance of climbing mountains or descending into craters and canyons, Marzwell asserts.

Mission-ready hoppers in the coming years would probably be smaller and lighter than today’s experimental models. They would also be equipped with cameras and one or two scientific instruments, each no larger than a sugar cube. “Like a colony,” Marzwell says, all the individual units would keep in touch with a “mother brain” on the orbiter or on one large rover on the surface.

If hoppers stay simple and cheap, “the return on investment for science will be 1,000 times what it is today,” Marzwell predicts. More than $100 million was spent on the lander and rover in the Pathfinder mission, he explains, whereas hoppers dropped from an orbiter would probably cost considerably less than $100,000 apiece and carry out more extensive scientific investigations.

What’s more, Marzwell argues, the penalties-financial and scientific-from losing a few of the robots would be minimal. “If I lose one or two, it’s not a big deal. The mission can go on,” he says.

While both of the critter-inspired robots’ designs leap beyond what roboticists have done before, they’re not the first efforts to make machines hop.

The record for the largest, heaviest hopper ever conceived may go to a leaping battle tank patented in 1945 by Henry W. Wallace. Although there’s no record that Wallace ever built the machine, he described it in detail in his patent application. He proposed a multiton metal canister bristling with cannons and bounding around on a single, telescoping leg. A driver would steer the leg, powered by explosives and diesel fuel, as it hurled the tank into the air with each thrust.

The idea of hoppers reappeared in a different guise when the space race began. Translated from German, a 1959 book called The Moon Car (Hermann Oberth, Harper) featured a hopping lunar vehicle.

In the 1960s, H.S. Seifert of Stanford University set to work on a hopping car that he intended to transport astronauts around the moon’s surface. He built a prototype and gathered evidence that hoppers are a particularly efficient mode of transportation in low gravity.

Robotic hoppers first appeared in the late 1970s. They were built by scientists interested in both robotics and new ways to study how people and animals walk and run. Those researchers built machines with sophisticated mechanisms and controllers that mimicked the ways animals balance and propel themselves.

Until this work, self-propelled robots were “like tables with moving legs,” says Marc Raibert, a pioneer of hopping and running robots who now heads a company called Boston Dynamics in Cambridge, Mass. During the 1980s and 1990s, Raibert and other scientists at Carnegie Mellon University in Pittsburgh, the Massachusetts Institute of Technology, and elsewhere developed a zoo of robots that hopped, ran, and even did flips.

The machines now under development in California and New Mexico represent a new type of hopping robot because they don’t maintain continuous motion or balance. Practical considerations ruled out that kind of sophistication. Says Robinett, “We’ve tried to make them as cheap, as dumb, and as simple as possible.”

Neither machine has yet debuted in a field application, but each one’s performance to date suggests where it might excel.

The Sandia researchers have, so far, created the faster jumper-leaping every 15 seconds or so. Their machine also lasts through many more hops on its stored energy.

The California design, however, may prove to be the more capable navigator because of two new features in its latest version. The frogbot now can control its jumping angle and can drive the last centimeters to its target on recently added wheels, Burdick says.

Compared with Sojourner, which roamed within only a few meters of its lander, the new generation of hopping machines looks toward much broader horizons-in space or on Earth.

“There is nowhere [this technology] can’t reach,” Marzwell declares. “The question is how many leaps will it take.”

Hoppers may keep minefields lethal

Antivehicle mines scattered from aircraft form an effective barrier against tanks and other military vehicles. However, if enemy troops blow up just one lane of the devices, a dangerous force can rumble through the breach.

Clearing a lane would become much harder if the remaining mines immediately flowed into the gap, much as water on a table, if you draw your finger through it, says condensed-matter physicist Thomas W. Altshuler of the Defense Advanced Research Projects Agency (DARPA) in Arlington, Va. He heads the agency’s program to develop a “self-healing minefield.”

The program is exploring the possibility of mounting 2-kilogram mines-shaped like half a coffee can-on the hopping robots invented by Sandia National Laboratories of Albuquerque. Alternatively, tiny rocket thrusters built into the mines’ rims might flip neighboring mines into the breach “like tiddlywinks,” Altshuler says. Whichever propulsion method is eventually chosen, the mobile mines will also require rudimentary intelligence and the ability to communicate with each other.

By fall 2002, DARPA aims to field-test systems of 50 such mines. Scientists are developing algorithms for determining how mines can efficiently move to fill gaps. “Those are very hard questions” that become even more challenging in actual minefields with hundreds or thousands of mines, Altshuler says.

Current U.S. military tactics call for sprinkling antipersonnel mines among antivehicle mines to deter enemy troops from clearing the antivehicle weapons. Because of those tactics, the United States has not signed the 1997 Ottawa Convention banning mines designed specifically to kill or maim people.

Following a directive from then-President Clinton, DARPA is developing the fluidlike minefield plan and other technologies to “obviate the need for the antipersonnel mine” and clear the way for U.S. endorsement of the treaty, Althshuler says.

More Stories from Science News on Tech