Date of Award

Spring 2022

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Mechanical Engineering & Materials Science (ENAS)

First Advisor

Dollar, Aaron

Abstract

Traditional manipulation planning and modeling relies on strong assumptions about contact. Specifically, it is common to assume that contacts are fixed and do not slide. This assumption ensures that objects are stably grasped during every step of the manipulation, to avoid ejection. However, this assumption limits achievable manipulation to the feasible motion of the closed-loop kinematic chains formed by the object and fingers. To improve manipulation capability, it has been shown that relaxing contact constraints and allowing sliding can enhance dexterity. But in order to safely manipulate with shifting contacts, other safeguards must be used to protect against ejection. “Caging manipulation,” in which the object is geometrically trapped by the fingers, can be employed to guarantee that an object never leaves the hand, regardless of constantly changing contact conditions. Mechanical compliance and underactuated joint coupling, or carefully chosen design parameters, can be used to passively create a caging grasp – protecting against accidental ejection – while simultaneously manipulating with all parts of the hand. And with passive ejection avoidance, hand control schemes can be made very simple, while still accomplishing manipulation. In place of complex control, better design can be used to improve manipulation capability—by making smart choices about parameters such as phalanx length, joint stiffness, joint coupling schemes, finger frictional properties, and actuator mode of operation. I will present an approach for modeling fully actuated and underactuated whole-hand-manipulation with shifting contacts, show results demonstrating the relationship between design parameters and manipulation metrics, and show how this can produce highly dexterous manipulators.

Download
COinS