迈向全自主超轻型无人机

Mihir Bala CMU-CS-25-111April 2025 Computer Science DepartmentSchool of Computer ScienceCarnegie Mellon UniversityPittsburgh, PA 15213 Thesis Committee:Mahadev Satyanarayanan, ChairDavid O’HallaronJeff SchneiderPadmanabhan Pillai Submitted in partial fulfillment of the requirementsfor the degree of Doctor of Philosophy. Copyright © 2025Mihir Bala This material is based upon work supported by the U.S. Army Research Office and the U.S. Army Futures Com-mand under Contract No. W519TC-23-C-0003 and by the National Science Foundation under grant number CNS-2106862. The content of the information does not necessarily reflect the position or the policy of the governmentand no official endorsement should be inferred. This work was done in the CMU Living Edge Lab, which is sup-ported by Intel, Arm, Vodafone, Deutsche Telekom, CableLabs, Crown Castle, InterDigital, Seagate, Microsoft, theVMware University Research Fund, IAI, and the Conklin Kistler family fund. Any opinions, findings, conclusions or recommendations expressed in this document are those of the author and donot necessarily reflect the view(s) of their employers or the above funding sources. Keywords:Autonomous Drones, Robotics, Mobile Computing, Edge Computing Abstract Autonomous drones have emerged as an exciting new technology which couldrevolutionize infrastructure inspection, military reconnaissance, and police surveil-lance. However, the vast majority of today’s platforms are heavy, costly, and difficultto operate. This restricts them from use in many mission settings, such as in denselypopulated environments, where government regulation forbids autonomous opera-tion of heavy drones near people.Much of this weight comes from the onboardcompute resources required for these drones to run the critical computer vision algo-rithms that provide situational awareness. In this dissertation, I show how autonomycan be induced on lightweight drones using edge computing, offloading high com-pute jobs to a network-proximal server. I demonstrate how this technique can leadto autonomous aircraft that fly much closer to the FAAs regulatory limits at accept-able performance cost. I also reveal a new operating system designed to unify thedisparate landscape of drones under a single, easy-to-program API. I show how thiscan be leveraged to create heterogeneous collaborative drone swarms on commercialoff-the-shelf hardware. Acknowledgments I would like to thank my family, especially my mom and dad, for their con-tinued support.I’d also like to thank my Pittsburgh friends: Bhavani Iyer, BrianSinger, Clement Fung, Abby Kroon Fung, Trevor Kann, Madeleine Howell Kann,Eric Zeng, Taro Tsuchiya, Joseph Reeves, Archana Iyer, Keane Lucas, Mark Dong,Alejandro Cuevas, and Amelia Cuevas; my college friends: Matthew Qemo, EvanArora, Matthew Askar, Jack Dolan, Lindsay Tuttle, Alimul Miah, Jagjot Singh,Tony Muscat, James Sabella, Sneh Bhakta, and Kelton Bursch; my DC friends:Alhan Sayyed, Adaah Sayyed, Lavan Rajan, Kiki Sayyed, and Nirmal Maitra; andlastly, my high school friends: Liam Wilson, David Schindler, Patrick Turiano, GrantKeith, and Alex Sutherland. Contents 1Introduction11.1Thesis Statement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21.2Thesis Overview. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .3 2Background and Related Work2.1The Development of Modern Drones . . . . . . . . . . . . . . . . . . . . . . . . 52.1.1Anatomy of a Drone. . . . . . . . . . . . . . . . . . . . . . . . . . . .72.1.2Regulation in Civilian Airspace. . . . . . . . . . . . . . . . . . . . . .92.2The Current COTS Drone Market. . . . . . . . . . . . . . . . . . . . . . . . .92.3What is Holding Back Drones? . . . . . . . . . . . . . . . . . . . . . . . . . . .102.4Prior Research on Autonomous Drones. . . . . . . . . . . . . . . . . . . . . .132.4.1Weight. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .142.4.2Accessibility. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .142.4.3Versatility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .142.4.4Portability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .152.4.5Cost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15 3Towards Better Autonomous Drones17 3.1The Advent of Edge Computing. . . . . . . . . . . . . . . . . . . . . . . . . .173.2Autonomous Drones and the Edge. . . . . . . . . . . . . . . . . . . . . . . . .183.3SteelEagle: Inducing Autonomy on Lightweight Drones. . . . . . . . . . . . .193.4Design Goals of SteelEagle . . . . . . . . . . . . . . . . . . . . . . . . . . . . .203.5Initial Hardware . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .213.5.1Control Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .223.5.2Networking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .223.5.3Video Stream . . . . . . . . . . . . . . . . . . . . . . . . .