Files in this item

FilesDescriptionFormat

application/pdf

application/pdfNILLES-DISSERTATION-2020.pdf (10MB)
(no description provided)PDF

Description

Title:Designing boundary interactions for simple mobile robots
Author(s):Nilles, Alexandra Q
Director of Research:LaValle, Steven M
Doctoral Committee Chair(s):LaValle, Steven M
Doctoral Committee Member(s):Amato, Nancy M; Mitra, Sayan; Murphey, Todd D
Department / Program:Computer Science
Discipline:Computer Science
Degree Granting Institution:University of Illinois at Urbana-Champaign
Degree:Ph.D.
Genre:Dissertation
Subject(s):robotics
minimalism
billiards
mobile robotics
boundary interactions
robophysics
live-coding
design
CAD
motion planning
underactuated robots
Abstract:Mobile robots are becoming increasingly common for applications such as logistics and delivery. While most research for mobile robots focuses on generating collision-free paths, however, an environment may be so crowded with obstacles that allowing contact with environment boundaries makes our robot more efficient or our plans more robust. The robot may be so small or in a remote environment such that traditional sensing and communication is impossible, and contact with boundaries can help reduce uncertainty in the robot's state while navigating. These novel scenarios call for novel system designs and novel system design tools. To address this gap, this thesis presents a general approach to modeling and planning over interactions between a robot and boundaries of its environment, and presents prototypes or simulations of such systems for solving high-level tasks such as object manipulation. One major contribution of this thesis is the derivation of necessary and sufficient conditions of stable, periodic trajectories for "bouncing robots," a particular model of point robots that move in straight lines between boundary interactions. Another major contribution is the description and implementation of an exact geometric planner for bouncing robots. We demonstrate the planner on traditional trajectory generation from start to goal states, as well as how to specify and generate stable periodic trajectories. In addition, we demonstrate the utility of the planner for environment geometry analysis, with respect to the role of environment geometry and system design constraints on the reachability and stability of bouncing robot trajectories. We propose a general approach for the design of bouncing robot systems, as well as more general classes of wild robots. We start by identifying useful, robust open-loop motion strategies, then integrate sensors and information space reasoning to determine conditions for switching between these open-loop behaviors. This approach is demonstrated on the task of "cart-on-track" object manipulation, motivated by design constraints for robots at the micrometer length scale. We identify the parts of this approach most amenable to automation, and provide a collection of supporting software tools. The final contribution of the thesis is a chapter including qualitative design principles for automated robot design systems, as well as an example of a live-coding interface for the design of mobile robot motion patterns demonstrating some of these principles. We conclude with an outline of our vision for this area of research in the future.
Issue Date:2020-12-02
Type:Thesis
URI:http://hdl.handle.net/2142/109428
Rights Information:Copyright 2020 Alexandra Nilles
Date Available in IDEALS:2021-03-05
Date Deposited:2020-12


This item appears in the following Collection(s)

Item Statistics