Vision-Aided Pose Estimation of an Object Manipulated by a Multifingered Hand. Application to Manipulation Control by a Soft-Fingered Hand.

摘要

Accurate object pose measurement is very important for manipulating objects by robotic hands. In this paper, we propose a systematic method to estimate the object pose with a complementary use of camera images and the finger joint information. Statistical characteristics of measurement errors are considered when two disparate sensor measurements are fused. The object pose is estimated not only at each camera sampling but also at inter-camera-sampling, so that the estimation can be used in the manipulation control servo loop.To verify the proposed pose estimation method, experiments are carried out. The proposed pose estimation method was combined with a control scheme for object manipulation by a multi-fingered hand with soft fingertips, which has also been proposed by the authors. Results of the experiment show the effectiveness of the proposed pose estimation method and the overall control system for object manipulation by a soft-fingered hand.