From Motor Babbling to Purposive Actions: Emerging Self-exploration in a Dynamical Systems Approach to Early Robot Development
Ralf Der and Georg Martius
Simulation of Adaptive Behavior 2006 (SAB 2006)
Rome, Italy, 25-29 September 2006
Summary
The results presented in the present paper can be considered as a step towards autonomous early robot development meaning the scenario where an unbiased robot might learn the essential sensorimotor coordination by self-exploration. The important point of our approach is that it is completely domain invariant, so that the emerging behaviors are dictated by the physical properties of the body and the environment. This has a direct bearing for embodied AI in the sense that our controller learns to excite certain physical modes of the body which are qualified by the fact that they can be understood by the model in easy terms. Hence we may understand these modes as behavioral primitives which may be used in more complex behavioral architectures.
We have given a theoretical approach to the cognitive deprivation problem which arises in the interplay between the world model and the controller. The system does not have any information on the structure and dynamics of the body so that the world model has to learn this from scratch. This involves the so called cognitive bootstrapping problem meaning that on the one hand the controls is to be such that the world model is provided with the necessary information. On the other hand these actions require a certain knowledge of the reactions of the body -- information is acquired best by informed actions. The concerted manner by which both the controller and the world model evolve during the emergence of the behavioral modes seems to be a good example of this process. ...