The human hand is commonly viewed as a near-optimal end effector — the envy of the robotics field. This thesis addresses the possibility of extracting some of its admirable qualities with respect to its actuation and applying them to robotics. Such a task is made possible through the use of models which emulate particular human functions.
Human muscle contraction is characterized by a number of nonlinear relations which can be described mathematically. These mathematical relations are appropriately combined into the design of a feedback actuation system. In this way human characteristics are used to provide the actuation for a single joint. The resulting model involves two muscle-like actuators which act antagonistically on either side of the joint. The system can be easily extended to actuate a model with any number of joints.
A simplified model of the human finger is designed, after thorough investigation of its function. The 3-link, 4-dof model involves six actuators which are configured to replicate the actions of the muscles and tendons of the human finger as well as that of one important ligament. The joint actuation system is applied to the finger model in order to generate the motion. This step is not carried out to perfection because of the shear size of the task involved, which was discovered in the process. A scheme of control is proposed for the finger model which would emulate actual human movement patterns. A set of six motion primitives are used as the basis for the control scheme as opposed to the more typical individual joint control.
The single-joint actuation model and the finger model are simulated using SIMNON, simulation software for nonlinear systems, on an IBM PC AT compatible computer.