A library for dynamic regrasping methods

Abstract

In this work we present a library of methods to perform dynamic regrasping to alternate grasp configurations of an object with respect to the task to be done. In these methods, we utilize the dynamics of the arm and therefore can use a simple non-dexterous gripper, enabling fast manipulations and low costs. We investigated two regrasping methods. One is an in-hand orienting approach where the gripper changes its orientation relative to a falling object. We have developed a novel stochastic motion planning algorithm that can be used for the in-hand regrasping manipulation under kinematic and dynamic constraints. The second is termed swing-up regrasping where an object is manipulated using a robotic arm around a point pinched by the arms gripper. The pivot point is modeled as a semi-active joint where only dissipative torques can be exerted and can only resist the motion of the object. We utilize an impulse-momentum approach to swing up the object and then a Clipped Linear Quadratic Regulator (cLQR) controller to stabilize the object in its desired angle. We also introduce the use of energy control for swinging-up the object.