Fingertip position and force control for dexterous manipulation through accurate modeling of hand-exoskeleton-environment

Despite mechanical advancements in the design of hand exoskeleton devices to help people with hand disabilities regain partial hand function, their manipulation performance has remained far inferior compared to the human hand. State-of-the-art control strategies implemented on exoskeletons are mainly focused on robot joint-level position control, although accurate control of fingertip positions and forces is a requirement for reaching human-like dexterity and manipulation. The relationships between inputs (motor commands) and outputs (fingertip positions and forces) are highly nonlinear due to the inherent limitations in actuation structure of multiple degree of freedom (DOF) exoskeletons. Moreover, the simplified coupled models of finger joint movements do not hold when humans interact with external objects and exert forces at their fingertips. Therefore achieving dexterous manipulation will require accurate models of interaction between the fingers, hand exoskeleton system, and fingertip environment.

In this thesis we accomplish, for the first time, fingertip position and force control with an assistive multi-DOF hand exoskeleton through accurate modeling of the hand-exoskeleton-environment. First, we provide kinematic and kinetic models for the human fingers, robot structure, and the Bowden cable power transmission for a fully actuated hand exoskeleton design. Next, we validate the models in simulation and demonstrate the successful control of fingertip position and forces in everyday drawing tasks. Finally, we utilize an experimental setup with a finger exoskeleton unit with two actuated DOF attached to an instrumented testbed finger to demonstrate successful tracking of fingertip position and forces within human accuracy levels through model-based control.