An Optical Tactile Sensor Assuming Cubic Polynomial Deformation of Elastic Body

Abstract

Assuming that the elastic body makes cubic polynomial deformation, we propose a compact three-dimensional (3D) optical tactile sensor for high-speed detection of three-axial directional force components. We constructed a 3D tactile sensor using thin, soft elastic and without pattern delineation or pigment injection such as that used in light-section measurement but having wider dynamic ranges and higher resolution. Conventional light-section measurement irradiating light onto sheets to measure objects requires a huge construction of the optical tactile sensor. Light-emitting diode (LED) sources are arranged around thin, deformable elastic membrane to obtain 3D force components from two-dimensional (2D) camera images taken using light sources of a minimum number of depth layers. Using two LED light sources - red and blue - around an elastic body, we estimate an object contact point pressing the elastic body and force magnitude and force incidence angle based on a mapping relationship predetermined through neural network learning from four ellipsoids formed by light irradiation and major and minor axis intersection points. To confirm that the elastic body forms cubic polynomial concavities at the point to rubber edges where force is applied based on X-, Y-, and Z-axes force components, we photographed elastic deformation and fitted curves into cubic polynomial expressions to investigate fitting accuracy. Fitting accuracy confirmed that cubic polynomials may reasonably approximate elastic deformation. We found that fitting curves onto cubic polynomials required two intersection points in addition to each edge of contact point and each of rubber edge point. Two is the minimum number of light sources required for irradiation. Experiments with this optical tactile sensor confirmed it to be effective in accurately estimating 3D elastic deformation, the object contact point, force magnitude, and force incidence angle.