Early cognitive development in learning autonomous control systems

Abstract

The author considers processes of early cognitive development (ECD) in learning autonomous control systems (LACS). The similarities and differences existing at the stage of ECD of humans and LACS are stated. It is shown that the two types of learning typical for LACS, i.e. quantitative (QL) and conceptual (CD), require different algorithmic structures. The roles of QL and CD in ECD are discussed. Examples from animated nature are considered, which imply the possible structure of systems for ECD in various LACS. The concept of the baby robot is devised as a tool for investigating ECD processes. This is an automaton moving autonomously in a state space of a problem under consideration, learning the state space, the problem, and the rules of operation. The baby robot is expected to be instrumental in the analysis of the mechanisms of autonomous generation of knowledge bases. The structure of a baby robot is considered to be a seed structure for building up self-organizing cognitive processes in robotics. >