Object manipulation of a humanoid robot based on visual Servoing

Abstract

Vision is a very important noncontact sensor for humanoid robots. In this paper, a method combining visual feedforward and visual feedback is proposed to implement reach-to-grasp task for a humanoid robot. Visual feedforward facilitates the reach-to-grasp task and reduces the manipulation time. Visual feedback increases the robustness by compensating the weak calibration error. The combination of two control strategies facilitates reach-to-grasp task for a humanoid robot. The robustness of the system is confirmed by the experiment results.