Embodied Cognitive Experiments: Applications to on-line navigation learning through teleoperation

It has been argued that robots cannot have full autonomy, as animals and humans can. Through learning, however, robots become more autonomous, in a technical sense of the term, to cope with their environments. Teaching is an efficient way for human learning, specially in basic learning, upon which further improvement can be acquired. To experiment with these ideas in robotics, we have developed a teaching scheme for robots to learn elementary navigation tasks on-line through teleoperation. This is accomplished by means of a neural network embedded in a robotic platform specially designed to handle embodied cognitive experiments. Possibilities in relation to a cognitive synthetic methodology, and also from a technical framework, will be presented.