Towards creating a flexible shape senor for soft robots

Abstract

Recent advances in robotics have witnessed an increasing transition from designing conventional robots with rigid components to partially or completely soft ones. Soft robots are known to be highly deformable and stretchable which makes the process of registering their shape and orientation in 3D challenging. This paper presents a first step of creating a flexible shape sensor for soft robots and a calibration algorithm that can compensate for different planar deflection conditions. In this paper, we describe the design and fabrication of the proposed shape sensor prototype utilizing three segmented optical fibers along the length of a flexible continuum arm. Three experimental scenarios of deflection are investigated to validate the relation between a mechanical deflection of the prototype and the change in intensity of the optical fibers' tip outputs (15 degrees deflection to the right and left, and planar double-bending). Camera images of the intensity circles without bending are used as a reference to relate the images features (location, angles, size, and intensity) to other bending cases. This study demonstrates the potential of relating the deflection status of a soft sensor to the image samples collected through a camera for the purpose of reconstructing and calibrating the shape sensor in 2D-space using MATLAB image processing toolbox and machine learning.