Evolving variable stiffness fiber patterns for multi-shape robotic sheets

摘要

Thin, planar sheets can be programmed to morph into complex shapes through stretching and out-of-plane bending, with applicability to shape-shifting soft robots. One way to make a morphing sheet is to use variable stiffness fibers that can modulate their tensile stiffness attached to the surface of a volumetrically expanding sheet. Adjusting local stiffnesses via tensile fiber jamming during sheet expansion allows control of the local shape tensor. However, finding the fiber placements and jamming policies to achieve a set of desired shapes is a non-trivial inverse design problem. We present an additive inverse design framework using an evolutionary algorithm to find optimal jamming fiber patterns to match multiple target shapes. We demonstrate the utility of our optimization pipeline with two input curvature pairs: 1) cylinder and sphere curvatures and 2) simple saddle and monkey saddle curvatures. Our method is able to find a diverse set of sufficient solutions in both cases. By incorporating hardware constraints into our optimization pipeline, we further explore the transfer of evolved solutions from simulation to reality.