Robotic Networks of Soft Linear Actuators

ORAL

Abstract

We consider the design of high elongation pneumatic linear actuators and control methods for robots made up of many of these linear actuators connected at universal joints. The result is a robot with natural compliance that is capable of dramatically changing its size and shape. The actuators are pneumatically powered and composed primarily of flexible but inextensible materials. To control robots formed by connecting many such actuators at universal joints, we derive the differential kinematics that relate changes in the actuator lengths to changes in the positions of the actuator endpoints, and show that controllability of the robot is equivalent to the infinitesimal rigidity of the underlying graph. Control methods are developed for two applications: locomotion and shape morphing. The control algorithm in both cases greedily minimizes an objective function while ensuring physical feasibility. For locomotion, the objective function is the motion of the robot’s center of mass along a prescribed trajectory. For shape morphing, the objective function is the distance between the robot's surface and a target shape represented by a point cloud. We present simulation results for both capabilities.

Presenters

  • Nathan Usevitch

    Mechanical Engineering, Stanford University

Authors

  • Nathan Usevitch

    Mechanical Engineering, Stanford University

  • Zachary Hammond

    Mechanical Engineering, Stanford University

  • Mac Schwager

    Aeronautics and Astronautics, Stanford University

  • Allison Okamura

    Mechanical Engineering, Stanford University

  • Elliot Hawkes

    Mechanical Engineering, University of California, Santa Barbara