Abstract
There are various manipulation methods of carrying large and heavy objects, and humans use these methods properly for each object. A robot needs to decide which manipulation method to use autonomously in order to perform without human help. In our method, a robot detects the motion state of the object by manipulating them with applying impedance control to hands. By manipulating the object, a robot gets kinematics and force information through hands touching the object. A robot can select a proper carrying method from pushing, pivoting, and dual-arm holding based on tilt / slide detection of objects motion and operational force limit detection. The effectiveness of our method is confirmed in an experiment in which a humanoid robot carries large furniture autonomously.
