ebook img

Planning Universal On-Road Driving Strategies for Automated Vehicles PDF

141 Pages·2018·4.129 MB·English
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 Planning Universal On-Road Driving Strategies for Automated Vehicles

AutoUni – Schriftenreihe H e in r ich Steffen Heinrich Planning Universal On-Road Driving P la n n in g U Strategies for n iv e r s a l O n - Automated Vehicles R o a d D r iv in g S t r a t e g ie s f o r A u t o m a t e d V e h ic le s AutoUni – Schriftenreihe Band 119 Reihe herausgegeben von / Edited by Volkswagen Aktiengesellschaft AutoUni Die Volkswagen AutoUni bietet Wissenschaftlern und Promovierenden des Volks- wagen Konzerns die Möglichkeit, ihre Forschungsergebnisse in Form von Mono- graphien und Dissertationen im Rahmen der „AutoUni Schriftenreihe“ kostenfrei zu veröffentlichen. Die AutoUni ist eine international tätige wissenschaftliche Einrich- tung des Konzerns, die durch Forschung und Lehre aktuelles mobilitätsbezogenes Wissen auf Hochschulniveau erzeugt und vermittelt. Die neun Institute der AutoUni decken das Fachwissen der unterschiedlichen Geschäftsbereiche ab, welches für den Erfolg des Volkswagen Konzerns unabding- bar ist. Im Fokus steht dabei die Schaffung und Verankerung von neuem Wissen und die Förderung des Wissensaustausches. Zusätzlich zu der fachlichen Weiterbildung und Vertiefung von Kompetenzen der Konzernangehörigen, fördert und unterstützt die AutoUni als Partner die Doktorandinnen und Doktoranden von Volkswagen auf ihrem Weg zu einer erfolgreichen Promotion durch vielfältige Angebote – die Veröffentlichung der Dissertationen ist eines davon. Über die Veröffentlichung in der AutoUni Schriftenreihe werden die Resultate nicht nur für alle Konzernangehörigen, sondern auch für die Öffentlichkeit zugänglich. The Volkswagen AutoUni offers scientists and PhD students of the Volkswagen Group the opportunity to publish their scientific results as monographs or doctor’s theses within the “AutoUni Schriftenreihe” free of cost. The AutoUni is an international scientific educational institution of the Volkswagen Group Academy, which produces and disseminates current mobility-related knowledge through its research and tailor-made further education courses. The AutoUni's nine institutes cover the expertise of the different business units, which is indispensable for the success of the Volkswagen Group. The focus lies on the creation, anchorage and transfer of knew knowledge. In addition to the professional expert training and the development of specialized skills and knowledge of the Volkswagen Group members, the AutoUni supports and accompanies the PhD students on their way to successful graduation through a variety of offerings. The publication of the doctor’s theses is one of such offers. The publication within the AutoUni Schriftenreihe makes the results accessible to all Volkswagen Group members as well as to the public. Reihe herausgegeben von / Edited by Volkswagen Aktiengesellschaft AutoUni Brieffach 1231 D-38436 Wolfsburg http://www.autouni.de More information about this series at http://www.springer.com/series/15136 Steffen Heinrich Planning Universal On-Road Driving Strategies for Automated Vehicles Steffen Heinrich Wolfsburg, Germany D issertation, Freie Universität Berlin, Department of Mathematics and Computer Science, 2017 Any results, opinions and conclusions expressed in the AutoUni – Schriftenreihe are solely those of the author(s). AutoUni – Schriftenreihe ISBN 978-3-658-21953-6 ISBN 978-3-658-21954-3 (eBook) https://doi.org/ 10.1007/978-3-658-21954-3 Library of Congress Control Number: 2018940659 © Springer Fachmedien Wiesbaden GmbH, part of Springer Nature 2018 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Printed on acid-free paper This Springer imprint is published by the registered company Springer Fachmedien Wiesbaden GmbH part of Springer Nature The registered company address is: Abraham-Lincoln-Str. 46, 65189 Wiesbaden, Germany Acknowledgements FirstandforemostIwouldliketothankmyadviserProf. Dr. RaúlRojas. Iwant to express my gratitude for giving me the opportunity, tools and trust to pursue researchinthefieldofroboticsasayoungundergraduatestudent.Hissupportand attitudetocontinuouslychallengethestateofthearthashelpmetosuccessfully finishthisthesis. IwasveryluckytospendmytimeasaPhDstudentattwouniqueplacesforauto- matedvehicleresearch. Mysincerethanksalsogoestomysupervisorandadviser atVolkswagenGroupResearch,Dr. ArneBartelsandDr. SimonGrossjohann,for theirguidance,supportandthoughtfuladviceonautomatedvehicles. Ithankmy mentor Dirk Langer and supervisor Jörg Schlinkheider at Volkswagen Electronic ResearchLabinBelmont,California,aswellasmycolleaguesJakeAskeland,So- mudroGuptaandJasonHardyforagreattimeofpioneeringautonomousdriving technology. IspecificallywanttothanktheresearchteamsatAutoNOMOSLabsinBerlinand VolkswagenAutomotiveInnovationLab(VAIL)inStanfordfortheirsupportand their valuable input as experts. In particular, I am grateful to Prof. Dr. Daniel Göhring,DanielSeifertandTinoschGanjinehforbeingmyearlyadvisersandgreat teamleads. I thank Prof. Dr. Raúl Rojas, Dr. Arne Bartels, Jannes Stubbemann and André Zoufahlforbeingco-authorsofmypublications. IwanttothankmyfellowPhDstudentsRobert,Bennet,Tobias,Daniel,Davidand Patrick for our countless round tables and reviews. I thank my students André, Sebastian, Alex, Jorit, Christian, Jannes and Tobias for their excellent research. I amgratefulforallfeedbackprovidedbyPatrick,Jens,Peter,Fritz,Tobias,Joritand Christianduringthefinalstagesofthethesis. MostimportanttothepursuitofmyPhDstudyhasbeentheconstantsupportof myfamily. Iwouldliketothankmyparentsfortheirencouragementanduncon- ditionalsupportaswellaslettingmetinkerwithtechnologyearlyon. Andmost importantly,Ithankmywifeanddaughterfortheirpatience,loveandoptimism duringthischallengingtime. SteffenHeinrich Table of contents ListofFigures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . IX ListofTables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . XI 1 Introduction,motivationandstructureofthethesis . . . . . . . . . . . . 1 1.1 Motivationofthethesis . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Thesisresearchquestions . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.3 Thesiscontributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Thesisoutline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 2.1 Introductiontomotionplanning . . . . . . . . . . . . . . . . . . . . . 7 2.2 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 2.3 Taxonomyofplanningmethods. . . . . . . . . . . . . . . . . . . . . . 11 2.4 Motionplanningforautomatedcars . . . . . . . . . . . . . . . . . . . 14 2.4.1 Diversityofdrivingenvironments . . . . . . . . . . . . . . . . 15 2.4.2 Planningassessmentcriteriaanddrivingmodes . . . . . . . . 15 2.5 Problemstatement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 3 Relatedwork . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1 Fromadvanceddriverassistancesystemstowardsautomatedcars . 21 3.1.1 Trajectoryplanning . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.1.2 On-roadswervepathgeneration . . . . . . . . . . . . . . . . . 25 3.1.3 Collisioncheckingandavoidance . . . . . . . . . . . . . . . . 26 3.2 Optimizationmethods . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.3 Drivingmodeselection . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 3.3.1 Multi-layeredsearchspacerepresentations . . . . . . . . . . . 31 3.3.2 Hierarchicaldrivingmodestatemachine . . . . . . . . . . . . 31 3.3.3 End-to-endmachinelearningapproaches . . . . . . . . . . . . 32 4 Aframeworkforuniversaldrivingstrategyplanning . . . . . . . . . . . 33 4.1 Planninginhighdimensionalstatespace . . . . . . . . . . . . . . . . 33 4.1.1 Identifyingkeycomponentsofsamplingbasedplanning . . . 35 4.1.2 PSPintegrationintoexistingarchitecture . . . . . . . . . . . . 37 4.1.3 Challengesandopportunities:Modelingauniversaldrive . . 40 4.1.4 Non-functionalrequirements:safety,comfortandacceptance 42 4.2 PSPworldrepresentations . . . . . . . . . . . . . . . . . . . . . . . . . 43 4.3 Visualizinghighdimensionalsolutions . . . . . . . . . . . . . . . . . 45 5 Sampling-basedplanninginphasespace. . . . . . . . . . . . . . . . . . . 49 5.1 Schematicofacompleteplanningsequence . . . . . . . . . . . . . . . 49 VIII Tableofcontents 5.2 Statespacesetupandexploration . . . . . . . . . . . . . . . . . . . . . 50 5.2.1 Statepropagationstrategies . . . . . . . . . . . . . . . . . . . . 51 5.2.2 Vehiclemotionguidedsampling . . . . . . . . . . . . . . . . . 52 5.2.3 Randomstatesampling . . . . . . . . . . . . . . . . . . . . . . 53 5.3 Trajectoryplanning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 5.3.1 Generatingpathsegmentswithclothoids . . . . . . . . . . . . 53 5.3.2 Generatingvelocityprofiles . . . . . . . . . . . . . . . . . . . . 57 5.4 Trajectoryoptimization . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 5.4.1 Extendedplanninghorizonwithmultiplelayers . . . . . . . . 60 5.4.2 Directedgraphoptimizationusingdynamicprogramming . . 64 5.4.3 OtherPSPoptimizationmethods . . . . . . . . . . . . . . . . . 69 5.5 Rulesandheuristics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69 5.5.1 Simplificationofthestatespace. . . . . . . . . . . . . . . . . . 71 5.5.2 Howtochooseaplanninghorizonlength? . . . . . . . . . . . 72 6 Auniversalapproachfordrivingstrategies . . . . . . . . . . . . . . . . . 75 6.1 Modelingdrivingbehaviors:Costfunctions. . . . . . . . . . . . . . . 75 6.1.1 Statetransitioncosts . . . . . . . . . . . . . . . . . . . . . . . . 75 6.1.2 Optimizingforauniversaldrivingexperience . . . . . . . . . 79 6.2 Situationawareness:Observingthemostrelevantthings . . . . . . . 82 6.2.1 Optimizingthevehicle’ssensorcoverage . . . . . . . . . . . . 83 6.2.2 Interpretationofasensorcoveragecostmap . . . . . . . . . . 87 6.3 Simulationexperiments . . . . . . . . . . . . . . . . . . . . . . . . . . 88 6.3.1 Validationofdrivingstrategygeneration . . . . . . . . . . . . 90 6.3.2 Evaluationofsmartpositioning . . . . . . . . . . . . . . . . . 93 6.3.3 PSPperformanceanalysis . . . . . . . . . . . . . . . . . . . . . 96 7 Modelingegomotionuncertainty . . . . . . . . . . . . . . . . . . . . . . . 103 7.1 Whymodelinguncertaintymatters . . . . . . . . . . . . . . . . . . . . 103 7.2 Egomotionuncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . 104 7.2.1 Apost-processingplanningextension . . . . . . . . . . . . . . 105 7.2.2 Stateestimationandmodelcorrection . . . . . . . . . . . . . . 109 7.2.3 Re-evaluatetrajectorycollisionprobability . . . . . . . . . . . 110 7.2.4 Collisiondetectionalgorithm . . . . . . . . . . . . . . . . . . . 112 7.3 Simulationexperiments . . . . . . . . . . . . . . . . . . . . . . . . . . 113 8 Summary,outlookandcontributions . . . . . . . . . . . . . . . . . . . . . 117 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 A Supplementalmaterial. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 List of Figures Figure1.1 Thesisoutline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Figure2.1 Pointmassrobotrepresentation. . . . . . . . . . . . . . . . . . . . 9 Figure2.2 Recedinghorizonmethod . . . . . . . . . . . . . . . . . . . . . . . 10 Figure2.3 Motionfeasibility . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Figure2.4 Configurationspacecelldecomposition . . . . . . . . . . . . . . . 12 Figure2.5 Vehicleautomation:Problemstatement . . . . . . . . . . . . . . . 14 Figure3.1 CharacteristicsofanEulerspiral . . . . . . . . . . . . . . . . . . . 26 Figure4.1 Systemsarchitecture . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Figure4.2 DiagramofthePSParchitecture. . . . . . . . . . . . . . . . . . . . 35 Figure4.3 Sampling-basedplanningkeycomponents . . . . . . . . . . . . . 36 Figure4.4 PSPframeworkinputdata . . . . . . . . . . . . . . . . . . . . . . . 38 Figure4.5 Integrationinexistingarchitecture . . . . . . . . . . . . . . . . . . 39 Figure4.6 Vehiclesensorset . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Figure4.7 Characteristicsofauniversaldrive . . . . . . . . . . . . . . . . . . 41 Figure4.8 Hierarchicaldecisionmaking . . . . . . . . . . . . . . . . . . . . . 42 Figure4.9 Differenttypesofcoordinateframes . . . . . . . . . . . . . . . . . 45 Figure4.10 PSPmodularvisualization . . . . . . . . . . . . . . . . . . . . . . . 46 Figure4.11 Visualizationofthesearchspace . . . . . . . . . . . . . . . . . . . 46 Figure5.1 SchematicflowforplanninginPSP. . . . . . . . . . . . . . . . . . 50 Figure5.2 Conesampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 Figure5.3 PSPsamplingseedpool . . . . . . . . . . . . . . . . . . . . . . . . 57 Figure5.4 Velocityprofilesforstatetransitions . . . . . . . . . . . . . . . . . 59 Figure5.5 Velocityprofileswithdifferenttimelimits . . . . . . . . . . . . . . 60 Figure5.6 PSPmulti-layerstructure . . . . . . . . . . . . . . . . . . . . . . . 61 Figure5.7 Temporalstatesampling . . . . . . . . . . . . . . . . . . . . . . . . 62 Figure5.8 PSPsearchgraphcomplexityreduction . . . . . . . . . . . . . . . 65 Figure5.9 PrincipleofDynamicProgramming . . . . . . . . . . . . . . . . . 65 Figure5.10 GPUmemorymanagement . . . . . . . . . . . . . . . . . . . . . . 67 Figure5.11 CUDAcostsimplementation . . . . . . . . . . . . . . . . . . . . . 69 Figure5.12 Localsamplingmethods . . . . . . . . . . . . . . . . . . . . . . . . 71 Figure5.13 Freespacebands . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 Figure5.14 Dynamicobstaclepredictions . . . . . . . . . . . . . . . . . . . . . 72 Figure5.15 Routing:Laneadvise . . . . . . . . . . . . . . . . . . . . . . . . . . 73 Figure6.1 Collisionchecking . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 Figure6.2 Maneuvercombination. . . . . . . . . . . . . . . . . . . . . . . . . 81 Figure6.3 Behaviormodeling:Doublelanechange . . . . . . . . . . . . . . . 82 X ListofFigures Figure6.4 Entropygrid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 Figure6.5 Positioningentropy . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 Figure6.6 Sensorcoverage . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 Figure6.7 VTDsimulationloop . . . . . . . . . . . . . . . . . . . . . . . . . . 89 Figure6.8 Drivingmaneuver:Start,followandstop . . . . . . . . . . . . . . 91 Figure6.9 Drivingmaneuver:Doublelanechange . . . . . . . . . . . . . . . 92 Figure6.10 Drivingmaneuver:Lanechange . . . . . . . . . . . . . . . . . . . 92 Figure6.11 Drivingmaneuver:Takingturns . . . . . . . . . . . . . . . . . . . 93 Figure6.12 Informationgainscenarios. . . . . . . . . . . . . . . . . . . . . . . 95 Figure6.13 AIEandFVCAIplotsforscenario1 . . . . . . . . . . . . . . . . . 95 Figure6.14 AIEandFVCAIplotsforscenario2 . . . . . . . . . . . . . . . . . 96 Figure6.15 Plot:CPUvs.GPUperformance . . . . . . . . . . . . . . . . . . . 97 Figure6.16 PSPhotspotanalysis . . . . . . . . . . . . . . . . . . . . . . . . . . 99 Figure6.17 PSPGPUperformanceoptimization . . . . . . . . . . . . . . . . . 100 Figure7.1 Problemstatement:Positionandmotionuncertainty . . . . . . . 104 Figure7.2 Egomotionuncertainty . . . . . . . . . . . . . . . . . . . . . . . . 105 Figure7.3 LGQpostprocessingpipeline . . . . . . . . . . . . . . . . . . . . . 106 Figure7.4 LQGinputdata . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Figure7.5 Scalingfactorck . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 Figure7.6 Scalingfactorinterpretation . . . . . . . . . . . . . . . . . . . . . . 112 Figure7.7 LQG-methodintegrationintoADTF . . . . . . . . . . . . . . . . . 113 Figure7.8 Visualizationofcollisionprobabilities . . . . . . . . . . . . . . . . 114 Figure7.9 Motionuncertaintydrivingscenarios . . . . . . . . . . . . . . . . 115

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.