Planning quasi-static fingertip manipulations for reconfiguring objects

Abstract

We address the global motion planning aspects of dexterous manipulation by a multifingered robotic hand. The specific task we address is: starting from a given initial grasp of a three-dimensional (3-D) object O, find feasible quasistatic trajectories (rolling/sliding motions and forces) for the fingertips to move O to a desired final configuration. We call this the reconfiguration problem. Our planner is based on a two-level algorithm combining a graph search on the configuration space of the object and a local planner that solves for instantaneous quasistatic motions of the entire manipulation system. The planner is used for simulating several complex reconfiguration tasks for smooth objects demonstrating the promise of our approach.