ebook img

System Design and Real-Time Guidance of an Unmanned Aerial Vehicle for Autonomous ... PDF

161 Pages·2015·7.14 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview System Design and Real-Time Guidance of an Unmanned Aerial Vehicle for Autonomous ...

System Design and Real-Time Guidance of an Unmanned Aerial Vehicle for Autonomous Exploration of Outdoor Environments Dissertation zur Erlangung des akademischen Grades Dr. rer. nat. an der Fakulta¨t fu¨r Mathematik, Informatik und Naturwissenschaften der Universita¨t Hamburg eingereicht beim Fachbereich Informatik von Benjamin Adler aus Hamburg November 2014 Gutachter: Prof. Dr. Jianwei Zhang Prof. Dr.-Ing. H. Siegfried Stiehl Tag der Disputation: 27. Februar 2015 Acknowledgements Completing this PhD project and writing this document took almost 4 years. During thistime, thecharacteroftheprojectoftenchanged, allowingmetoworkwithdifferent people, read literature, create software and integrate hardware. As such, it was a very interesting and rewarding time, bringing about many failures as well as a few achievements. First of all, I would like to thank my advisor Prof. Dr. Jianwei Zhang for being a constant source of guidance, support and criticism throughout these years. Also, Prof. Dr.-Ing. H. Siegfried Stiehl kindly accepted to review this thesis, taking a lot of time and giving helpful guidance in the process. Thank you, Prof. Stiehl! Furthermore, I would like to express gratitude to my colleagues at TAMS, who pro- vided a very productive environment in which ideas can evolve from “ridiculous!” to “works!”. When everyone is keen to listen, laugh and help, this makes for a great place to learn. Houxiang, thank you for giving me a great start! Special thanks go out to you, Bernd, for your generous support, incredible patience and thoughtful advice. Also, thank you for always having time to hold the fishing rod! Junhao, I am grateful for both our friendship and collaboration. Because of you, I got to see another culture, learn a tiny bit of Chinese history and eat a lot of great food! Benjamin Adler v Acknowledgements vi Abstract The central endeavor of this thesis is the successful development and evaluation of an aerial mobile robotic system that produces precise, georeferenced, three-dimensional maps of outdoor environments by means of autonomous exploration. The system consists of a ground station and a custom-built UAV with six degrees of freedom, featuring an on-board computer, an inertial navigation system, and two 2D laser range finders. In addition to a description of the hardware architecture and individual components being used, this dissertation presents challenges and problems that arose during the construction of its hard- and software as well as optimizations applied in the course of its development. Afundamentalpartofthisworkisthedistributedsoftwarearchitectureforin-flightsen- sor fusion and data analysis, with a focus on a novel, truly three-dimensional algorithm generatingmultiplenext-best-views(NBVs). Designedforapplicationonairborneplat- formsinoutdoorenvironments, theapproachworksdirectlyonraw, unstructuredpoint clouds and can be used either indoors or outdoors with any sensor generating spatial occupancy information. Based on the generated sensor-poses and the incrementally growing point cloud, tra- jectories are computed for the UAV to autonomously map its environment. To ensure safe operation, collision avoidance constantly monitors the planned path and updates it whenever obstacles are detected. In order to satisfy real-time constraints, all algorithms are implemented on a highly parallel SIMD architecture found in modern GPUs, allowing for extremely fast motion planning and responsive visualization. As the underlying hardware imposes limitations with regards to memory access and concurrency, necessary data structures and further performance considerations are explained in detail. Data has been captured during real, autonomous flights and is used to analyze the performance of all major components (flight controller, next-best-view generation, dy- namic path planning and collision avoidance) of the system in realistic outdoor sce- narios. The performance of the GPU-based next-best-view algorithm is also compared against a previous, CPU-based proof of concept. vii Abstract viii Kurzfassung Die vorliegende Dissertation berichtet u¨ber die Entwicklung eines luftgestu¨tzten Robo- tersystems zur autonomen Erkundung von Außenumgebungen mit dem Ziel der Erstel- lung pr¨aziser, georeferenzierter Karten. DasvorgestellteSystembestehtauseinerBodenstationsowieeinereigenskonstruierten Drohne,welcheeinenComputer,einNavigationsystemundzweiLaserscannermitfu¨hrt. Neben einer Beschreibung der Hardwarearchitektur und ihrer Komponenten werden Herausforderungen und Einschr¨ankungen bei Systemintegration und Softwareentwick- lung, sowie Optimierungen im sp¨ateren Verlauf der Entwicklung pr¨asentiert. Ein zentraler Teil der Arbeit ist eine verteilte Softwarearchitektur zur Fusion und Ana- lyse von Sensordaten im Flug, wobei hier der Fokus auf einem neuartigen Algorithmus zur Erstellung von next-best-views aus bereits erstelltem Kartenmaterial liegt. Obwohl zur Anwendung auf luftgestu¨tzten Plattformen zur Kartographie von Außenumgebun- gen mittels LIDAR erdacht, ist der Algorithmus direkt auf jegliche ra¨umliche Sen- sordaten (z.B. von s¨amtlichen Laserscannern, RGBD- und Time-of-Flight-Kameras) anwendbar. Basierend auf den erzeugten Sensorpositionen und dem aktuellen Stand der Karte erzeugt das System sichere Flugrouten zur weiteren Kartierung des Gebiets und u¨berwacht diese auf Hindernisse. Um einen Einsatz in Echtzeit zu erm¨oglichen, sind die Algorithmen auf parallel arbei- tenden GPUs implementiert. Dies erm¨oglicht nicht nur schnelle Flugplanung, sondern auch effiziente und interaktive Visualisierung der aufgenommenen Daten und der Ar- beitsweisedereingesetztenAlgorithmen. DadiezugrundeliegendeHardwarebesondere Anforderungen an Speicherzugriffsmuster und Nebenla¨ufigkeit stellt, werden Daten- strukturenundweitereU¨berlegungenzurLeistungssteigerungimDetailerla¨utert. Weiterer Bestandteil der Arbeit ist eine Analyse von Daten aus echten Flu¨gen in Außenumgebungen zur Beurteilung der Anwendbarkeit und Leistungsf¨ahigkeit aller verwendeten Komponenten (Flugsteuerung, Erzeugung von next-best-views, Pfadpla- nungundKollisionsvermeidung). FernerwirddasLaufzeitverhaltendesGPU-basierten next-best-view Algorithmus mit dem eines CPU-basierten Prototyps verglichen. ix Kurzfassung x

Description:
Data has been captured during real, autonomous flights and is used to analyze the .. NAVSTAR Navigation System using Timing and Ranging ToF Time of Flight. TSP Traveling Salesman Problem. UAS Unmanned Aircraft System. UAV Unmanned Aerial Vehicle. UTM Universal Transverse Mercator.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.