Perception-based robotics based on perceiving-acting cycle with modular neural networks

Abstract

Discusses controlling and learning methods of perception-based robotics from the viewpoint of ecological psychology. In the method, the perceptual system and the action system of the robot restrict each other according to the spatio-temporal context of the environment. Furthermore, we discuss the unit of actions using modular neural networks for robotic systems. We apply the proposed method to a dynamic environment. Several computer simulations show the robot can learn behaviors based on the perceiving-acting cycle.