How is the robot taught to perform its task?

Programming of most current robots is performed using basic instructions such as:

• Raise control arm to index (find out where you are)
• Open manipulator fingers to index (find out where you are)
• Turn motor shaft clockwise to index point (find out where you are)
• Turn motor shaft counter-clockwise 35 pulses of the encoder. (correlates with a specific angle)
• Lower control arm until resistance is detected on manipulator 'down' pressure switch.
• close manipulator fingers until fingertip pressure detector detects resistance.
• Raise control arm to index.
• turn motor shaft clockwise an additional 70 pulses of the encoder...

These types of basic motions will vary from one robot to another, and can be grouped together into "routines", allowing the engineer to "call" a specific task rather than program the entire set every time.

Robots do not learn by watching someone perform the task.

Another way that a robot can "learn" a task would be for the engineer to physically move a robot through a specific task, using something like a "joy stick" and instructing the robot to store the moves for later repetition; but in that case, the robot would simply be storing instructions similar to what is described above.