Hybrid Control of Multi-fingered Dextrous Robotic Hands

Abstract

This article presents a hybrid discrete-continuous dynamical systems approach to control of multi-fingered dextrous manipulations by robotic hands. A discrete-continuous modeling framework allows to derive a hybrid state model integrating time-driven dynamic features of manipulation systems as well as discrete event aspects resulting from varying contact situations between the robotic hand and manipulated objects. The combination of continuous and discrete dynamic aspects is typical for multi-fingered manipulation. A hybrid control architecture comprising a hybrid planning scheme for grasping and regrasping, impedance control algorithms based on 6D contact force sensor information, and a formal compensation method for discrete contact state errors are proposed. Results of dynamical simulations and experiments with a four-fingered hand grasping and manipulating objects demonstrate the robustness improvement in grasping control as achieved by use of the proposed hybrid control approach.

The work reported here was performed while the first and second authors stayed with the Institute of Automatic Control Engineering, Technische Universität München, Germany. Experiments were conducted at the Tokyo Institute of Technology, Japan.