Implementation of an adaptive neural controller for sensory-motor coordination

摘要: A theory and prototype of a neural controller called INFANT, which learns sensory-motor coordination from its own experience, is presented. INFANT adapts to unforeseen changes in the geometry physical motor system location, orientation, shape, size objects. It can learn accurately grasp an elongated object without any information about system. relies on self-consistency between sensory signals achieve unsupervised learning. designed be generalized for coordinating number inputs with limbs joints. implemented image processor, stereo cameras, five-degree-of-freedom robot arm. After learning, average position accuracy within 3% length arm, orientation 60 degrees solid angle. >