Image‐based backbone reconstruction for non‐slender soft robots

Abstract

AbstractOne advantage of soft robots to conventional robots is their continuous deformation, which makes them highly adaptable to their surroundings. To describe the kinematics of soft robots it is not sufficient to only know the position of the end‐effector. Instead, the steady state of a soft robot can be described by its backbone curvature. An image‐based backbone reconstruction method is used and adopted for a non‐slender soft robot to reconstruct the backbone curve using polynomials to describe an arc length‐dependent curvature. A constant curvature and cubic curvature polynomial have been tested to reconstruct the backbone curve. In the case of a pressurized soft robot without externally applied loads, the constant curvature polynomial can be used to reconstruct the backbone of the soft robot using an image‐based approach, yielding good agreement with the captured images. However, some care has to be put into the consideration of the soft robot geometry in this method, to guide the optimization procedure. From our findings, we can conclude that it is not necessary to apply markers to the soft robot to reconstruct the backbone, nor is it necessary to have a detailed geometric model of the soft robot to be used in a differential rendering process. An image‐based iterative‐closest‐point optimization method, with some carefully placed estimate points, to integrate the volumetric structure of the soft robot can yield good results for the reconstructed backbones.