Learning motor primitives for robotics
ICRA'09 Proceedings of the 2009 IEEE international conference on Robotics and Automation
From neural networks to the brain: autonomous mental development
IEEE Computational Intelligence Magazine
Hi-index | 0.00 |
Learning how to control arm joints for goal-directed reaching tasks is one of the earliest skills that need to be acquired by Developmental Robotics in order to scaffold into tasks of higher Intelligence. Motor Babbling seems as a promising approach toward the generation of internal models and control policies for robotic arms. In this paper we propose a mechanism for learning sensory-motor associations using layered arrangement of Self-Organizing Neural Network (SOINN) and joint-egocentric representations. The robot starts off by random exploratory motion, then it gradually shift into more coordinated, goal-directed actions based on the measure of error-change. The main contribution of this research is in the proposition of a novel architecture for online sensory-motor learning using SOINN networks without the need to provide the system with a kinematic model or a preprogrammed joint control scheme. The viability of the proposed mechanism is demonstrated using a simulated planar robotic arm.