Fostering Techniques for Cognitive and Motor Development

The degree of the human controls robot behaviors tends to polarize toward the extremes. At one end the human is in charge of: controlling the robot as a marionette, or feeding its brain with everything one assumes the robot should know.

At the other hand, the robot is seldom provided with a set of learning algorithms and left alone in the world to learn by built maps, exploration, make sense of it by itself. This is very challenging job, and in many respects we throw the robot to the lions, e.g. the danger in the real, unpredictable world.

The midway is to have a continuous, active involvement of a human during the development of the set of capabilities the robot needs in the world. In the fostering of animal world is considered an important component to ensure survival of the species. It is been observed that the more ”advanced” a species, the longer the period of immaturity of its offspring, in other words the longer the parents need to foster their children. It is the period when the young ones develop the skills that would make them successful in life.

Humans learn by themselves or from others. In the initial learning phase they may learn movements with out any control or information from others, while later they may learn under total guidance and control. When the learner does not know what controls to give to the muscles, he can learn by exploration, children learned their first movements this way, for instance, during learning eye-hand coordination they flail randomly their hands and record the perceptions they get for the applied controls, associating perceptions with actions.

Robots could learn in the same ways what humans do. Robots learned the control of sensory motor by exploration, following the circular reaction mechanism. Exploration is one way how to generate examples of associations between perceptions and actions. The guidance can be done by analogic teaching, which is useful particularly when the precise coordinates where the robot should go are not known exactly, but the operator can see where he or she wants the robot to move.