Buckling instability for directional control in biomimetic soft robots

We report a data-driven method to control the locomotion of a bacteria-inspired soft robot by using buckling instability. The soft robot is composed of a spherical head and a helical elastic rod rotating in low Reynolds fluid. We use an experimentally validated fluid-structure interaction simulator that combines Discrete Elastic Rods algorithm with Resistive Force Theory. The robot follows a straight path below a threshold rotational speed. Buckling ensues in the rod at this threshold and the robot takes a nonlinear trajectory in its post-buckling phase. Even though the simulator can predict the trajectory, solving the inverse problem of following a given path simply by controlling the angular velocity poses a challenge for traditional analytical tools. This led us to adopt data-driven techniques from the machine learning community and exploit the robustness and speed of the simulator to use it as a data generator. We demonstrate that the soft robot can follow a prescribed path only by tuning its angular velocity. Our results may shed light on the original microscopic system of bacterial locomotion that inspired our robot design.