Positional estimation of soft actuators through embedded sensing

Abstract

Soft robotic systems are prone to bending in undesired modes due to the compliant materials used to fabricate them. These systems use pneumatic soft actuators relying on internal pressure to induce specific deformations. Controlling soft actuators requires system modeling, specifically in systems of multiple actuators. Acquiring accurate models through analytical methods is challenging due to embedded components constraining elastic deformations of the body. A less arduous approach to this challenge is observing system parameters through integrated sensing to develop an estimated model. As packaged soft sensing components are limited, research and development of the physical sensor and technique for implementation is required. Developing these methods involves mechanical design of components, electronic instrumentation, placement for target deformations, interpretation of response for embedded control, and benchmarking performance. The documentation of these attempts at sensor development serves to expand the toolbox of techniques for sensing in soft robotic bodies. Two approaches were made to the process. The first involved embedding conductive fluidic strain sensors in a bidirectional soft actuator of an assistive grasp prosthetic to characterize the undesired modes of deformation it was experiencing. Sensors were placed to observe the intended mode of bending as well as undesired rotational sag and twist of the actuator body. These sensors provided haptic feedback of actuator position that enabled users to correctly identify grasp in 83.3% of responses. The second was the design and fabrication of a soft grasp prosthetic incorporating control from embedded positional sensing. The primary grasping actuators were deployed with an embedded fiber optic array monitoring reflectance against an internal diaphragm. These sensor embedded actuators were employed in feedback control to independently alter grasp configuration to objects of varying shapes.