Optimization of an Underactuated Two Finger Robotic Hand Using Genetic Algorithms

In this work we focus on the design of an underactuated anthropomorphic tendon driven robotic gripper with two fingers. The gripper is required to be able to perform strong stable flat pinching and enveloping grasps for a fixed actuation force by maximizing the total contact force with the object. In order to design this gripper, we use a Genetic Algorithm (GA) based optimization approach, where geometrical parameters, such as position, stiffness and size of the joints, the length of each link and the palm, the distance from the tendons to the joint centers as well as the starting angle of the fingers are used. The approach runs the whole grasping process for each individual of the GA in simulation, detecting first constraint violations, and then measuring the contact forces. The optimization procedure was experimentally validated by 3D printing a prototype of the optimal design and showing its grasping capability.