Self-organizing model of learning and its application to eyesight system of robot

摘要

A self-organizing model that dynamically constructs balanced trees of a database to universally learn any relation between input signals and output signals of unknown systems, is proposed. Contrary to conventional databases that memorize data individually, the self-organizing model forms an associative memory which realizes the function of learning. Applications of the model to position identification and obstacle avoidance in an autonomous mobile robot system, as well as to continuous handwritten letter recognition are shown, together with some successful experimental results.