Date of Award
Spring 2023
Document Type
Dissertation
Degree Name
Doctor of Philosophy (PhD)
Department
Mechanical Engineering & Materials Science (ENAS)
First Advisor
Dollar, Aaron
Abstract
In-hand manipulation is a complicated process—it requires a group of individual manipulators to work in proper unison with one another, modulating forces on an object while making and breaking contact. Traditional approaches to realizing such capabilities have been through the use of overly-complex, anthropomorphic, and thus expensive, robot hands equipped with a vast array of sensors, which are often noisy and can in turn lead to task failure. Mechanically compliant end effectors, on the other hand, have shown to be beneficial to robot manipulation, particularly for grasping, as they are able to "absorb the slack” in any modeling, sensing, or control uncertainty. As grasping is a necessary element to in-hand manipulation, it can further be hypothesized that compliant end effectors could be similarly beneficial for extending these capabilities. However, motions of compliant mechanisms are typically complex and difficult to analytically model. Moreover, there remain questions in how to appropriately determine and represent system state—what features need to be tracked, how accurate do they need to be, and how often do they need to be captured? These questions can be studied through the lens of grasp mechanics via vision-based feedback, which can play a crucial role for such devices that often lack onboard sensing as to keep designs simple, compact, and inexpensive. Formally, I will present an approach that is able to observe the state of a compliant hand-object system during manipulation, control the object along a planned path for fixed-contact in-hand manipulation, and finally, desirably plan the trajectory of a grasped object within-hand given constraints. To close, I will extensively showcase the applicability of these methods in tight tolerance and open-world assembly tasks.
Recommended Citation
Morgan, Andrew, "Observing, Controlling, and Planning Compliant Robot Manipulation" (2023). Yale Graduate School of Arts and Sciences Dissertations. 1089.
https://elischolar.library.yale.edu/gsas_dissertations/1089
