Grasp Synthesis for Dextrous Hands Optimised for Tactile Manipulation

In this article we present a novel method for generating collision-free force-closure grasps optimised for applications that utilise tactile array sensors for the grasping task. In general the problem of computing good grasps for dextrous hands is complicated and time consuming. Popular approaches use simulation to generate databases of good grasps off-line, so when given task specific quality criteria, feasible grasps can be efficiently computed online. In this work we extend the grasp database approach to include dynamically simulated grasps with simulated tactile measurements. We furthermore show how to use this in an online grasp planner to find feasible grasps for tactile manipulation tasks. We demonstrate the method in a setup that includes the Schunk Dextrous Hand with 6 tactile array sensors with which we show simulated results.