| dc.description.abstract | Workers across multiple sectors of industry are often put at risk of developing musculoskeletal disorders such as carpel tunnel syndrome, tendinitis, and lower back pain. These work-related musculoskeletal disorders stem from repeatedly lifting heavy equipment, resisting tool vibrations, and exerting forces in non-ergonomic postures. While full automation could completely alleviate the issue, many industrial tasks such as equipment maintenance and repair require human sensory presence. For these scenarios, researchers have developed collaborative robots that can offset the physiological strain from the worker while still being safe enough to operate in close physical proximity to the worker. Despite their potential benefits, there are still areas of industrial work for which no collaborative robot has been designed. One such area is industrial operation inside confined spaces. Human-robot collaboration in confined spaces presents a host of difficult technical challenges to ensure worker safety during complex industrial tasks.
This dissertation aims to address a few of these technical challenges in the areas of design, sensing, and control. We first present the mechanical design of a robot intended for human-robot collaboration in confined spaces. This robot consists of a statically balanced, rigid-linked base, two continuum segments lined with contact and proximity sensors, and a wrist for a total of 11 active degrees of freedom. The static balancing of this robot is achieved using a spring-loaded wire-wrapped cam mechanism. When designing this mechanism, we noticed several gaps in the literature. To address these limitations, we present an optimization-based design strategy for two-degree of freedom wire-wrapped cam mechanisms that 1) respects the deflection limits of springs, 2) ensures the cams are convex, and 3) minimizes the effect of unmodeled changes in spring constant. We also present a model of the effect of wire-cam friction and evaluate the method experimentally. Next, we present what we believe is the first adaptation of the generalized momentum observer contact detection/estimation approach for variable curvature continuum robots. We also present a model for the effect of dynamic state uncertainty on the performance of the observer and evaluate the method experimentally. After this, we present a redundancy resolution and end-effector compliance modulation strategy for robots with bracing constraints. This redundancy resolution strategy can be used to enable the design of long-reach robots with minimal torque actuators. Lastly, we present the software and control framework of the aforementioned collaborative robot. We also present a preliminary system evaluation user study that explores the benefits and tradeoffs of human sensory presence for mock industrial tasks. Specifically, the user study explores controlling the robot using admittance control and using teleoperation with and without virtual fixtures.
We believe the contributions to the design, sensing, and control of collaborative robots in confined spaces presented in this dissertation will help improve the health of industrial workers and increase the adoption of collaborative robots. | |
