The latest of the robots have been programmed to mimic human actions when human are subjected to obstacles (done through motion capturing using sensors) and act accordingly when left to decide on their own before similar situations. For example, a robot learns to swing a baseball bat through motion capturing from a real human who is swinging a baseball bat facing a ball and is fitted with sensors at every joint and tips of the limbs. Then the robot is subjected to similar situation and the robot repeats the same action. But the robot follow exactly the same angle, speed and direction, whether or not the ball can be struck with that.

So far, the latest robot (the best one from the point of view of capable of taking decisions) can hardly do what an one and a half year old child does with ease. So, you can imagine the long way it has to go. It often tumbles and falls down while crossing three single bricks laid in its path. It easily crosses the first, takes on the second with difficulty and falls down on the third.

I cannot explain the learning algorithms here.

Asker's Rating:

Permalink | Report