Intelligent optimization based on the genetic algorithm for a customizable Stephenson III six-bar mechanical finger

Abstract

The motion of the hand’s fingers allows humans to perform many activities. A mechanical model of these limbs can be used in industry and healthcare applications. Due to the sophisticated structure of such limbs, the generation of mechanisms to emulate them is complex but can be addressed with computational intelligence techniques such as metaheuristics. Current models consist of closed, open, or hybrid kinematic chains. Each alternative has advantages and disadvantages in terms of cost, energy, precision, variety of movements, and anthropometric and anthropomorphic characteristics. These mechanisms are derived from information obtained from hand biomechanical studies or clinical experience, so they are not considered customizable and are hardly anthropometric and anthropomorphic. This work presents an approach for the intelligent synthesis of customizable mechanical fingers with anthropomorphic and anthropometric features. This approach aims to exploit the relatively low cost, high precision, and complex trajectories that can develop the one-degree-of-freedom Stephenson III six-bar mechanism to perform cyclic flexion and extension movements as a human finger would. For this, the dimensional synthesis problem of the six-bar mechanism is proposed as an optimization one. So, anthropometric characteristics of the finger are accounted for by using a reference trajectory derived from precise measurements of the subject’s cyclic flexion and extension movements relative to the metacarpophalangeal joint. On the other hand, anthropomorphic features are incorporated by imposing constraints that induce dimensions of the mechanism that resemble the human finger, regulate the size of the links corresponding to hand bones, and place fixed points in locations that mirror the metacarpal structure. The characteristics obtained through this approach have not been found in any design similar to this one to date. With the proper synthesis of the mechanism, it is intended to track an anthropometric reference trajectory collected from the finger of a healthy individual through a commercial low-cost optical hand sensor and conditioned using the spectral clustering unsupervised learning technique. This approach successfully synthesized a customized mechanical finger for a test subject using a genetic algorithm. The design was implemented through low-cost additive manufacturing. After several analyses, the proposal proved to be accurate in tracking the finger movements of different individuals, flexible to anthropometric data, and possessing advantages over other alternative metaheuristics approaches.