Drone Doggie Built for War
(DefenseTech) A robot dog could one day become a soldier’s best friend — if an Army program works out as planned. Today’s soldiers carry as much as 100 pounds of equipment. That’s exhausting, even for the toughest grunt. In the future, the Army wants to dump up to half that gear onto the back of a drone. But military scientists are worried that robots with wheels won’t be able to follow their human masters across mountain passes, up stairs and through forest trails.
To make their way across that kind of terrain, the drones will need legs — maybe even four of them. So the Army’s Tank-automotive and Armaments Command, or TACOM, has just doled out $2.25 million to two robotics firms to prototype a big, mechanical dog capable of carrying ammunition, food and supplies into battle.
The contracts are part of a broader Pentagon look into robots that take their cues from nature. Defense Department-backed scientists are studying swarms of bees and packs of wolves for ideas on how to get drones to work together. Man-made snakes, lobsters, flies — even elephant trunks — are just a few of the animal-inspired devices being created by military-funded researchers.
“We’re coming full circle,” said Paul Meunch, a TACOM research scientist. “In the days of George Washington, the Army used mules and horses. Then it moved on to trucks. And then armored vehicles and tanks. Now we could be swinging back to four legs.”
But reaching that galloping dream won’t be easy. Building mechanical legs that work right has been a brutal task. Spinning a wheel is simple. Swinging a set of legs that can bend, step high and keep a robot balanced is hard.
“We’re at the bottom of the pyramid right now,” said Ben Krupp, president of Yobotics, which won a $750,000, two-year TACOM grant to build a Great Dane-sized drone. “It’s tough just to get a four-legged robot to run across the parking lot without falling down.”
After decades of research, tiny, commercial robo-dogs can now scamper across a flat surface. Child-sized humanoid bots can waddle — carefully. A canine drone in the armed forces would have to do much better, though, keeping up with soldiers marching over uneven terrain.
Powering the dog will be one of the biggest barriers to bringing it into a war zone, Krupp said.
“You can strap an internal combustion engine on its back. But until battery technology catches up, you’re not going to have a stealthy robot, because it’ll be pumping out 110 decibels when it’s running down the street,” he said.
But robotic leg pioneer Marc Raibert said significant power can be saved by getting the dog to act more like an animal.
A computer brain telling a set of motors to spin every time a robot wants to take a step is not very efficient, and it’s not how animals work.
“The body has a mind of its own, with reactions that are governed by the physics of the situation. Getting harmony between the body, mind and the computer brain — that’s the real challenge of the project,” Raibert said. His company, Boston Dynamics, won a $1.5 million TACOM robotic canine contract recently — in addition to a grant from Darpa, the Pentagon’s research arm, for related work. He calls his creation the “Big Dog.”
Running involves more than swinging a leg, of course. There’s a cycle of small motions — a gait — that moves a body forward. And, in recent years, researchers have gotten a much better handle on how those tiny movements add up to something larger, Carnegie Mellon University professor Howie Choset said. The Office of Naval Research is funding his efforts to build a wheeled, mechanical elephant’s trunk for inspecting ships’ engines.
The military has become interested in the work of roboticists like Choset because their biologically inspired bots can go places wheeled robots — and maybe even soldiers — can’t. Lobsters can navigate choppy water and scurry across the ocean floor undetected. So the Navy is putting money into a robo-lobster to find mines along coastlines. Being a “fly on the wall” in an enemy hideout has always been a secret agent’s dream. So Darpa is backing research into mechanical insects.
Despite all this activity, Meunch said, getting the Army to pony up money for the drone dog has been tough.
“The Army’s been reluctant to fund it, so far. They’ve been very conservative, when it comes to funding legged research,” he said. “So this project is mostly a technology demonstrator, to get further investment by the Army.”
How well the four-legged robots perform will depend largely on how well they can see. The natural, intuitive process of recognizing obstacles and avoiding them may be easy for dogs and people. But for robots, it’s no simple task. Depth perception is particularly tough.
To solve this, Raibert has enlisted the help of Larry Matthies, who heads the machine vision group at NASA’s Jet Propulsion Laboratory. Matthies’ team has developed a way for robots to see in three dimensions, like humans and animals do. Two separate cameras take images of the same scene. Every pixel in the left camera is matched up with ones in the right camera. Then the two images are combined — and a sense of depth is created, just like in the old 3-D movies.
The Web-going public should be thoroughly familiar with Matthies’ work. His team at JPL built a similar system for NASA’s Spirit rover, now sending back spectacular images from Mars.
