Model-based design for full-stack robot manipulation

Abstract

As robotic manipulation research becomes more prevalent, it is crucial to develop clean, modular, and testable systems for researchers to make advances in their fields of expertise without sacrificing usability for others. It is also important to introduce those methods to beginners as they are entering the field, so they can build upon them in novel research. This thesis introduces a framework, informally called the Manipulation System, built using Drake Systems based on work done for the Fall 2018 MIT class Intelligent Robot Manipulation. This thesis also presents the groundwork for performing full-stack robot manipulation tasks and proposes extensions to make the system more usable and accessible to all.