Escape Behaviour for the Task-Level Control of a Soft Robot

Abstract

A soft robotic surface manipulator is a device equipped with a computer-controlled, deformable surface. By moving the surface, objects placed on the robot can be transported, rotated, or otherwise brought into a desired configuration. A probabilistic automaton can use camera feedback to measure and control the movement of the work pieces, and find the optimal sequence of actions to reach a desired state. The probabilistic nature of the method solves most problems resulting from noise and impulsive disturbances. In some cases, however, there are systematical disturbances that are out of the scope of a probabilistic method. In a soft robotic surface manipulator, this can lead to objects becoming pinned against an obstacle or performing repetitive sequences of actions. This paper describes a method to escape from these situations. The method is supported by simulations based on the measured properties of an existing soft table. This is the first publication dedicated to this problem in soft robotic, task-level control.