Sensory integration with articulated motion on a humanoid robot

Abstract

The work in this thesis seeks to integrate the motion of a humanoid robot with its auditory and visual sensory information to achieve various reflex actions that mimic those of people. Such reflexes in the form of reach-grasp behaviors can enable the robot to learn through experience its own state and that of the world. A humanoid robot with auditory capabilities, stereo vision, and artificial pneumatic arms and hands was used to demonstrate tightly coupled sensory-motor behaviors in five different demonstrations. The complexity of succeeding demonstrations was increased to show that the reflexive sensory-motor behaviors combine to perform increasingly complex tasks. The humanoid robot executed these tasks effectively and established the ground-work for the further development of hardware and software systems, sensory-motor vector-space representations, and coupling with higher level cognition.