Mechatronic design and genetic-algorithm-based MIMO fuzzy control of adjustable-stiffness tendon-driven robot finger

Abstract

This work presents a detailed design of a three-jointed tendon-driven robot finger with a cam/pulleys transmission and joint Variable Stiffness Actuator (VSA). The finger motion configuration is obtained by deriving the cam/pulleys transmission profile as a mathematical solution that is then implemented to achieve contact force isotropy on the phalanges. A VSA is proposed, in which three VSAs are designed to act as a muscle in joint space to provide firm grasping. As a mechatronic approach, a suitable type and number of force sensors and actuators are designed to sense the touch, actuate the finger, and tune the VSAs. The torque of the VSAs is controlled utilizing a designed Multi Input Multi Output (MIMO) fuzzy controller. The fuzzy controller input is the force sensors' signals that are used to set the appropriate VSA torque. The fuzzy controller parameters are then tuned using a genetic algorithm as an optimization technique. The objective function of the genetic algorithm is to avoid unbalance torque in the individual joints and to reduce the difference between the values of the supplied VSAs torques. Finally, the operation of the aforementioned finger system is organized through a simple control algorithm. The function of this algorithm is to enable the detection of the unknown object and simultaneously automatically activate the optimized fuzzy controller thus eliminating the necessity of any external control unit.