Including virtual constraints in motion planning for anthropomorphic hands

Abstract

This paper copes with the problem of finding a collision-free path for a hand-arm robotic system from an initial unconstrained configuration to a final grasping (or preshape) one. The aim is to obtain a natural motion as a sequence of human-like postures that both capture the coupling that there exist between the fingers of the human hand and also maintain the palm oriented towards the object to be grasped. The proposed method is a sampling-based approach whose efficiency relies in the reduction of the dimensionality obtained by considering, for the finger joints, a subspace determined by the main principal motion directions that capture the coupling and, for the position and orientation of the palm, the submanifold that satisfies the orientation constraint. The approach is illustrated with an example and compared to the case where no virtual constrains are used, validating the proposal.