INTRODUCTION TO UNDER-ACTUATION
The human hand is a remarkable gripper which roboticians have attempted to imitate for a long time. But, as we can imagine, we are far from achieving such an imitation. With its 20 degrees of freedom, 19 muscles, 17 joints, 19 bones, in addition to its ligaments, nerves and numerous sensors, the human hand is very difficult to reproduce mechanically. The Utah/MIT hand represents a good example of a robotic hand: despite its almost human appearance, this hand requires an arsenal of cables, electronic components and software to support its operation.
Between very complex robotic hands and simple grippers, it is possible to find very interesting grippers, including underactuated robotic hands.
Underactuated robotic hands are the intermediate solution between robotic hands for manipulation (versatile, stable grasps, expensive, complex control, many actuators) and simple robotic grippers (simple control, few actuators, task specific, unstable grasps). In an underactuated robotic hand, the number of actuators is less than the hand's degrees of freedom (DOFs). The mechanical intelligence embedded into the design of the hand allows the automatic shape adaptation of the fingers (see the figure bellow). The underactuated DOFs are governed by springs and mechanical limits. No special control is needed.
An example of the use of an underactuated 2-DOF finger is shown in Figure 2. The finger closes under the influence of a force represented by the red arrow. When the first phalange comes into contact with the object, the second phalange pivots until it enters into contact with the object. Note the presence of a spring and a mechanical limit at the second joint to maintain the stable position of the second phalange when there is no contact.
By grouping several fingers together it is possible to obtain a complete gripper, simple to control but capable of carrying out complex grasping tasks. The main possible grasps are illustrated in Figure 3.
The first three images show delicate grasps in three different orientations of the fingers. One can see a delicate cylindrical, spherical and pinching grasp. The last two images show englobantes grasps, including a cylindrical grasp and a spherical grasp. Note that for each type of grasp, the object does not have to have a regular shape; the hand automatically adapts to the shape of the object to be grasped. This is clearly demonstrated by our numerous prototypes of underactuated robotic hands.