In September of 2015 I successfully defended my thesis. I would like to thank my committee members, Drs. Volkan Isler, Stergios Roumeliotis, Eric Frew, Mihailo Jovanovic, and Arindam Banerjee for their inspirational work and helpful direction.
My thesis is available as:
J. Vander Hook. University of Minnesota 2015.
[permalink] [pdf] [bib]
Thanks to advances in miniaturization, computing power, reliable sensors, and battery life, mobile robots are increasingly being used for a wide variety of environmental monitoring tasks. No longer confined to factory floors or controlled environments, robots for remote sensing in dangerous or hard-to-reach environments could provide the same scalability, precision, and reliability to environmental monitoring as they did to industrial applications. To enable this kind of long-term, reliable, autonomous mobile sensor deployment, algorithms which can ensure that the robots achieve their sensing tasks are required.
In this dissertation, we present fundamental results in using mobile sensors to locate targets of interest. This Active Localization problem forms the core study of the thesis. The dissertation is roughly separated into three main parts.
In the first part of the thesis, we study the problem of using one or more mobile robots equipped with bearing sensors to locate a stationary target in minimum time. The problem requires optimizing the measurement locations of the robots to gather the required information about the target's location. In addition, when multiple robots collaborate, we include communication constraints in the path planning objective. Two formulations for this problem are studied. First, we study the offline problem of finding measurement trajectories when the true target location is known. Second, we study the online version and show how to adapt the offline solution to the situation when the target location is not known, while preserving the quality guarantees of the offline solution.
In the second part of the thesis, we study the problem of locating multiple stationary targets using a single mobile robot. We formulate a novel coverage problem and provide two main results. We first study the problem of initializing consistent estimate of the targets' locations. These initial estimates are used to seed an active localization algorithm which is shown to localize the targets quickly. In a second formulation, we assume that the targets are within a set of polygonal regions, but have no further information about the distribution or number of targets in the environment. An algorithm is provided which can choose measurement locations to localize all the targets to within desired precision in near optimal time.
In the third part of the thesis, we study the problem of using bearing information to track and capture a moving target. We present two formulations based on pursuit-evasion games. In the open plane, the objective is for a mobile robot to minimize the distance to a maneuvering target when only uncertain bearing information is available to the robot. Then, we study the problem of capturing the maneuvering target in a closed environment by moving close to it. We show that the size of the environment relative to the sensing noise determines if this is possible.
In addition to theoretical results, we present field studies of using one or more mobile robots to detect radio transmitters using these results. We show that the algorithms presented are suitable for use in monitoring invasive fish.
This dissertation provides both fundamental theoretical studies of active localization using bearing sensors and extensive field studies which verify the usefulness of the results in environmental monitoring tasks.
I thought when I finished this dissertation I would feel great. As in large or capable significantly above the average amount. Completing a large project like this over five years is not something to attribute to my own abilities. Any milestone is the sum of the individual decisions that led to it and when I look at the amount of guidance, encouragement, patience, and downright goading that it took to make the decisions that led to this point, I feel only gratitude.
Understanding that this list is incomplete, I first thank the person most influential to this work: my advisor, Dr. Volkan Isler. You reminded me to be calm under stress, to do good work first and find a place for it second, to treat those around me with the utmost respect, and to never use a colon to break a sentence. Thanks for taking a chance on a terrified first-year that walked into your office. I owe you much.
Thanks to my committee, Professors Roumeliotis, Banerjee, Frew, and Jovanovic, for your feedback before (and possibly after) my submissions for publication, your patience with my answers, challenges to my assumptions, and guidance and assistance in finding a place to grow now that my time here is over.
Thanks to my lab mates for dinners after experiments, sympathy when the equations would not work, feedback when my ideas were ridiculous, and help pulling the robot out of the muck or snow when it got lost. Thanks especially to Pratap, Patrick, Narges, and Haluk, with whom I lived and worked the closest. We publish singly but we suffer and celebrate in aggregate. I wish you the best. Remember, everyone wears the waders.
Thanks to my family and those I've added to it. Friends you make during the hard times are friends you make for life. For your patience, encouragement, much-needed distractions, perspective, and love, I'm grateful. Thanks, Jared, for brewing and inspiration. Chase, you got me through a tough year and I'm glad we could adventure. Thanks to Lisa, Andy, Pete and my boxing buddies at Uppercut for the bruises, blood, and confidence. Thanks to Hanna for hannalysis, Marie, Robert, Emery, and Ben for commiseration in the dungeon of Keller Hall, and the rest of the outstanding grad student population for all the happy hours. Thanks to Worms, Bergo, Andy, and Matt for teaching me to pile in and cleave. To Geoff, Bryce, Willard and the rest: skol. To Chrissy, a picture is worth a thousand words. I can't say it any better: Vamos.
And finally, thank you to the National Science Foundation, ARCS Foundation, University of Minnesota, Sigma Xi, and patient parents for providing financial support.