Motion planning for 3D multifingered caging with object recognition using AR picture markers

Abstract

Caging is a method to capture an object geometrically by position-controlled robots without any force and tactile sensors. Many previous researches focused on caging constraints of objects, and those on planning are few. In this paper, we present a motion planner for caging by a multifingered hand and a manipulator to produce whole motion which includes approaching to a target object and capturing it without any collisions. We derive sufficient conditions required for the caging tasks about three caging patterns. Since the planner requires the object properties including the position and orientation of the object, we adopt an object recognition using AR picture markers. We apply the proposed method to caging about four target objects: a cylinder, a ring, a mug and a dumbbell. Some experimental results shows that each motion are successfully planned, and executed by the arm/hand system.