Post-Grasp Pose Adjustment of a Tendon-Driven Underactuated Robotic Hand Using Finite-Action-Set Receding-Horizon Planning

Abstract

This paper addresses local post-grasp pose adjustment for a tendon-driven underactuated robotic hand. A finite-action-set receding-horizon motor reference planning method is proposed, in which the continuous motor reference increment is discretized into representative actions and an action-effect table predicts local object pose changes. At each decision instant, candidate action sequences are evaluated using pose error, action magnitude, motor limits, grasp margin, and action switching, while only the first action of the optimal sequence is executed. Simulations show that the method generates feasible commands for single-axis and two-dimensional pose adjustment, drives the object pose toward the target, and keeps motor references within the allowable range. Compared with one-step planning, a longer prediction horizon reduces action switching in the two-dimensional coupled task and improves action sequence smoothness.