The cute, childlike robot, named iCub, was designed by researchers at the Italian Institute of Technology. Armed with a bow, an arrow, a cute (if politically incorrect) Native American headdress and a complicated computer algorithm, the robot learns from his missed shots iteratively, until he makes the bull’s-eye.

The task of firing an arrow, the researchers say, was picked for its inherent and obvious reward, and simultaneous marriage of motor control with image processing. Nothing to do with arming a bunch of human-hating robots to the teeth, allegedly.

iCub uses a learning algorithm called ARCHER (Augmented Reward Chained Regression), which implements a camera to process the bull’s-eye image, and his previously fluffed attempts, to figure out the perfect angle, force and trajectory to make the winning shot.

The first iteration of iCub hit the bull’s-eye, standing three and a half metres from the target, in eight attempts. Here’s hoping the next few iterations don’t whittle it down to two or three trials while replacing the bow with a shotgun.

It’s the latest robotic creation at the technology institute in Italy that learns complicated tasks through a series of iterative trial and error attempts. Earlier this year, the same institute taught a Barrett WAM 7 robotic arm to flip pancakes. That one took a slightly more lengthy 50 trials to master.

The archery-mastering iCub will be presented at the Humanoids 2010 conference in Tennessee this December. According to the conference’s program, he’ll be joined by a passenger carrying biped, musical conducting robots, a Mini-Humanoid Pianist and a robot that can play table tennis.

Originally published on Wired UK.

Photo credit: Petar Kormushev/Wired UK

See Also:

• Robot + Super Gun = 'Crowd Control'
• Atlantan Builds Robot Vigilante
• Robot Cannon Kills 9, Wounds 14
• Wired 14.01: The 50 Best Robots Ever

Authors: Mark Brown, Wired UK

to know more click here