Contact Trajectories for Regrasp Planning on Discrete Objects

Abstract

Manipulation tasks, in general, require a grasp change on the object during its execution. The manipulation problem can be solved by simply rolling or sliding the fingers on the object surface, the so-called regrasping approach. This paper provides an algorithm for regrasp planning of 2D and 3D discrete objects, such that the regrasp trajectory of each contact ensures a force-closure grasp (i.e. a grasp that resists external disturbances) while the regrasp motion is performed. The proposed approach takes advantage of a sampling-based method that quickly explores the grasp space, and relies on the use of independent contact regions and non-graspable regions, which provide large regions of the force-closure or non force-closure subspaces starting from a single sample. Application examples are included to show the relevance of the results.