Design of a Compliant Underactuated Robotic Finger with Coordinated Stiffness

The concept of underactuation has been previously developed in the robotic field for grasping applications. For these anthropomorphic grippers, the minimization of the number of input signals, or in other words underactuation, is the most expected characteristic. This method has become very popular in recent decades. Indeed, by minimizing the number of input signals, it minimizes the complexity of the system’s control and at the same time avoids increased weight and cost. The inconvenience of such a technique is that the design of this type of system remains a difficult task if the behavior of the underactuated set of joints is to remain within certain constraints. In this thesis, an underactuated robotic finger whose grasping behavior is modulated by the design of its superelastic joints is proposed. Using shape-memory alloy, the finger joints can be given specific stiffness and pre-form shapes such that a single-cable actuation rather than opposing-pair actuation can be used; this also allows the grasping motions of the phalanges to be synchronized in the free phase and then adaptive once contact is made. A default-closed pre-tensioned configuration allows grasp forces to be maximal for larger objects and still keeps control components such as tendons out of the grasp workspace. The simplicity of the design lends itself to the possibility of integrated joint angle and surface pressure sensing on the finger itself. The details of design, prototyping and testing are described. Advisor: Carl A. Nelso