dc.description.abstract | The advantage of soft actuators and robots compared to their rigid counterparts is often articulated as the inherent compliance afforded by the hyper-elastic materials used to construct soft robots and the high power density of actuation using fluid power. Although seemingly intuitive, the manner by which hyper-elasticity and fluid compliance in these systems interact to affect actuator stiffness is generally not well understood, specifically in soft actuators with no volumetric changes with deflection. Literature suggests that in actuators with deflection dependent volumes, such as McKibben actuators, the actuation fluid stores elastic potential energy, and thus the bulk modulus determines the stiffness. However, for actuators not driven by a significant volume-deflection relation, such as bending inflated tubes and Pneu-Net actuators, there is a gap in understanding about how fluid pressure affects physical stiffness and elastic potential energy storage outside of empirical relations and complex, heavily constrained finite element models.
Compared to conventional motors, soft actuators may exhibit hard to predict dynamic behaviors due to their inherent compliance, making it difficult to design dynamic controllers. Modeling efforts, empirical solutions, and learning methods have been employed to control soft robots, but the lack of fundamental understanding of the physical properties of soft actuators produce complex and computational expensive controllers. Through informed design and descriptive modeling that is conducive to traditional control tools, simple and easy to implement controllers can be made.
In my dissertation, I aim to provide an understanding and application of the mechanisms that enable energy-storing, spring-like stiffness in fluid-filled soft actuators. Specifically, I focus on (i) constitutive modeling of stiffness in soft, pneu-net actuators (ii) the design of a variable stiffness, bi-directional pneu-net actuator and (iii) the physical impedance control of the soft actuator in a collision task. | |