This white, 41-inch-tall kid-sized robot with a Native American headdress is called iCub, and he learned how to shoot a suction cup-tipped plastic arrow all by himself at the Italian Institute of Technology.

The ARCHER (Augmented Reward Chained Regression) algorithm doesn’t memorize and recreate a demonstration of best angle and strength to hit the center of a bull’s eye target 3.5 meters away; rather, it’s only taught the rules of the game (hit the center of the target) and estimates better parameters based on the results of a previous try, which it observes with image recognition. The iCub was able to hit the center of the bull’s eye in only eight rollouts!

The research will be presented in December at the Nashville, Tennessee event Humanoids2010.