Toward vision-based adaptive configuring of a bidirectional two-segment soft continuum manipulator

Abstract

In soft robotics, developing an effective way of robot-environment interaction is a challenging task due to the soft nature of the material that makes the manipulator. This paper demonstrates a vision-based approach to configure a two-segment soft continuum robot manipulator into an user-defined configuration and interact with unknown objects on plane. The soft robot manipulator actuated by cable-driven mechanism, is composed of two cascade continuum segments which are made from poly-dimethyl-siloxane (PDMS). The overall robot configuration can be determined in a point-wise manner on image plane provided by an eye-to-hand system. One can define the end-effectors' location on the visual system to re-shape the manipulator. The visual servoing fashion allows the robot to optimize its posture to its best fit without developing any complicated model. Experiments on prototype indicate that the proposed model-free approach can be well employed, even when the manipulator is bearing a payload. By adaptively adjusting manipulator's stiffness to a quasi-deadlock status, the payload capacity is up to nearly 6 times of the manipulator's mass itself.