ROBO-​​PUPPIES GO TO SCHOOL

Darpa, the Pentagon’s mad sci­ence divi­sion, wants to teach lit­tle mechan­i­cal pup­pies to think. Hopefully, that’ll let the bots run around with sol­diers on the bat­tle­field one day.
Getting robots to maneu­ver around rocks and trees and pot­holes is tough — just ask any of the tin­ker­ers whose bots bit the dust dur­ing last year’s all-​​drone off-​​road rally across the Mojave Desert.
One way around the prob­lem, some drone-​​makers think, is to give their cre­ations legs, so that they can maneu­ver just like a per­son or an ani­mal would. But that’s eas­ier said than done. Walking, it turns out, requires a zil­lion tiny cal­cu­la­tions to keep bal­ance and avoid obsta­cles. It’s so com­plex, Darpa notes, that “hand­craft­ing the con­trol laws and para­me­ters” needed for robots to hike “may not even be pos­si­ble with rea­son­able effort.” So instead, Darpa would like to get the bots to fig­ure out how to walk on their own.

In the Learning Locomotion pro­gram, algo­rithms will be cre­ated that learn how to loco­mote based on the expe­ri­ence of a legged plat­form con­fronting extreme ter­rain. It is expected that the per­for­mance of these algo­rithms will far exceed the per­for­mance of hand­crafted sys­tems, cre­at­ing a break­through in loco­mo­tion over extreme ter­rain. Further, it is expected that these algo­rithms will be broadly applic­a­ble to the class of agile ground vehicles.

Darpa is plan­ning on hand­ing out a series of $600–800,000 con­tracts to try to teach drones to walk. And the robots the agency wants researchers to train are 6.6 pound, 10.6 inch-​​long “Little Dogs.“
During the 15-​​month first phase of the “Learning Locomotion” project, Darpa wants the pooches to be able to travel .6 of an inch per sec­ond, and scale obsta­cles about 2.5 inches tall. For Phase II, those num­bers should go up to approx­i­mately 3.8 inches and 5.7 inches, respec­tively.
That may not sound like much. Bu the drones will have to be smart enough that that can “learn ‘on-​​the-​​fly’ how to tra­verse new obsta­cle types,” Darpa tells researchers. “Government tests will mea­sure the abil­ity of the per­former sys­tems to learn from expe­ri­ence.“
This isn’t the only Darpa pro­gram to try to get ground-​​based bots to think. Nor is this the only Defense Department project which involves dog-​​like drones. Last year, the Army doled out $2.25 mil­lion to two robot­ics firms to pro­to­type a big, mechan­i­cal pooch capa­ble of car­ry­ing ammu­ni­tion, food and sup­plies into bat­tle.
THERE’S MORE: Oh, this rules.

Flesh and bone tri­umphed in the first ever man-​​versus-​​machine bat­tle of brawn — an arm wrestling con­test between robots and humans in California on Monday.
The cham­pion, beat­ing all three robotic arms each in mat­ter of sec­onds, was a 17-​​year-​​old girl called Panna Felsen, a high school stu­dent from San Diego, US.
The con­test was set up by Yoseph Bar-​​Cohen at NASA’s Jet Propulsion Laboratories in Pasadena, California, US, in an attempt to encour­age the devel­op­ment of polymer-​​based arti­fi­cial mus­cles… The ulti­mate aim is to have an arti­fi­cial arm beat the world’s strongest per­son, says Bar-​​Cohen. But for now he wanted to make the chal­lenge slightly more attain­able which is why Panna, a self-​​confessed wimp, was cho­sen to rep­re­sent human­ity.
Despite her lack of strength, train­ing and tech­nique, she was able to con­quer the first arm… in just 24 sec­onds. Following this, and a pep talk from an arm wrestling expert, it took her just four sec­onds to beat the sec­ond arm and three sec­onds for her to win the last match.

March 8th, 2005 | Drones | Comments Off Both comments and pings are currently closed.