Abstract
Stiffness control provides a mechanism for controlling finger position or force, and facilitates stable behavior during the transition between unconstrained motion and sudden contact with the environment. The method proposed here provides uniformity of response upon finger contact for any contact stiffness, as long as no separation occurs. The stiffness control system of a finger joint in a robot hand was partitioned into linear and nonlinear subsystems. The controller design used pole placement techniques based on the linear subsystem while the mechanical nonlinearities (i.e., load and velocity dependent nonlinear friction and nonlinear damping) in the drive were modeled separately. The parameters of the nonlinear model were experimentally identified off-line. These identified parameters were then used in a real-time estimator for compensation of the nonlinear effects while the system was under stiffness control. The technique was implemented successfully at 40 HZ on the actual finger under investigation. The results are a significant improvement on traditional techniques for nonlinear systems which result in large offsets or unstable behavior.

This publication has 0 references indexed in Scilit: