A Bio-Inspired Robotic Finger Driven and Shape-Sensed by Soft Optical Tendons

Abstract

Nature-inspired robotic manipulators are an active research area because of their efficient multi-functional integration. Of these mechanisms, the tendon-driven human finger is capable of sophisticated manipulation tasks but is challenging to implement as a robotic design. A key challenge is managing tendons for force transmission from muscle to fingertip within the limited space of the hand. Another challenge is controlling in underactuated tendon-driven systems. Sensor feedback is usually the answer, but an increase in sensor systems adds bulk, which is the opposite direction to develop for a dexterous design. Therefore in this article, a novel design of a human index finger-inspired robotic finger is presented. The finger is composed of three major parts that make up the functionality of the finger motion and detection. First, a rigid resin material for the bone structure. Second, a soft rubber resin for sheaths that guide tendons and set each joint's angular range. Third, the soft optical tendons that transmit force and sense the total joint angle of the finger through the light intensity. The correlation of the force and joint angle from the soft optical tendons measurements is further analyzed by neural network. Evaluation of the sensor reliability is demonstrated through image acquisition. The data shows that the maximum RMSE of training and testing session are 2.36(deg) and 5.69(deg) respectively.