Vinícius J. Cassol Soraia R. Musse Cláudio R. Jung Norman I. Badler Simulating Crowds in Egress Scenarios Simulating Crowds in Egress Scenarios í Vin cius J. Cassol Soraia R. Musse (cid:129) á Cl udio R. Jung Norman I. Badler (cid:129) Simulating Crowds in Egress Scenarios 123 Vinícius J.Cassol Cláudio R. Jung Virtual Humans SimulationLab TheInstitute of Informatics PontificalCatholicUniversityofRioGrande Federal University of RioGrandedoSul doSul Porto Alegre, RioGrande doSul Porto Alegre, RioGrande doSul Brazil Brazil Norman I.Badler SoraiaR. Musse Department ofComputer andInformation Computer Science Department Science PontificalCatholicUniversityofRioGrande University of Pennsylvania doSul Philadelphia, PA Porto Alegre, RioGrande doSul USA Brazil ISBN978-3-319-65201-6 ISBN978-3-319-65202-3 (eBook) https://doi.org/10.1007/978-3-319-65202-3 LibraryofCongressControlNumber:2017954876 ©SpringerInternationalPublishingAG2017 Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpart 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 orinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar methodologynowknownorhereafterdeveloped. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse. 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 authorsortheeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinor for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations. Printedonacid-freepaper ThisSpringerimprintispublishedbySpringerNature TheregisteredcompanyisSpringerInternationalPublishingAG Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland (Vinicius) For my dad Vicente and mom Liris, who taught me how beautiful life can be. (Soraia) To my parents (Eduardo and Iara) and daughters (Marina and Helena), who taught me the two first rules of love: respect and care. (Claudio) To my mom and dad (Ligia and Hilario), for all the care and incentive during my student years (and my whole life!!!). To my daughters Marina and Helena, for teaching me how to love. (Norman) To international colleagues everywhere. Contents 1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 2 Background Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.1 Theories in Crowd Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 2.1.1 Crowd Dynamics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.1.2 Crowd Evolution in Egress Situations. . . . . . . . . . . . . . . . 10 2.2 Regulations in Evacuation Processes . . . . . . . . . . . . . . . . . . . . . . 12 2.3 Crowd Simulation in Emergency Situations . . . . . . . . . . . . . . . . . 14 2.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3 Crowd Simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 3.1 Existing Technologies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 3.1.1 Crowd Control Trainer. . . . . . . . . . . . . . . . . . . . . . . . . . . 20 3.1.2 MassMotion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.1.3 Legion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 3.1.4 Massive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 3.1.5 Golaem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 3.1.6 Menge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 3.1.7 Miarmy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.1.8 Houdini Crowds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 3.1.9 LCrowdV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2 CrowdSim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 3.2.1 Setup and Simulation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 3.2.2 Simulation of Heterogeneous Agents . . . . . . . . . . . . . . . . 33 3.2.3 Graphical Visualization of Simulations . . . . . . . . . . . . . . . 34 3.3 CrowdSim Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 3.3.1 Component Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.3.2 Qualitative Validation . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 vii viii Contents 3.3.3 Quantitative Verification . . . . . . . . . . . . . . . . . . . . . . . . . 43 3.3.4 Observing Emergent Behaviors in CrowdSim . . . . . . . . . . 44 3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4 Case Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 4.1 Case Studies Pipeline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48 4.2 Evacuation of Olympic Stadium . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.3 College Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52 4.4 School . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55 4.5 Night Club. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 4.5.1 Simulation of Heterogeneous Agents Under Alcohol Influence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70 4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 5 Crowd Analysis Based on Computer Vision . . . . . . . . . . . . . . . . . . . 73 5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 5.2 People Counting and Density Estimation . . . . . . . . . . . . . . . . . . . 74 5.2.1 Pixel-Level Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 5.2.2 Texture Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 5.2.3 Object-Level Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 78 5.2.4 Datasets and Validation . . . . . . . . . . . . . . . . . . . . . . . . . . 80 5.3 Tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 5.3.1 Multiple Pedestrian Tracking . . . . . . . . . . . . . . . . . . . . . . 82 5.3.2 Crowd Flow Estimation . . . . . . . . . . . . . . . . . . . . . . . . . . 85 5.3.3 Datasets and Validation . . . . . . . . . . . . . . . . . . . . . . . . . . 86 5.4 Behavior Understanding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88 5.4.1 Unusual/Abnormal Event Detection . . . . . . . . . . . . . . . . . 89 5.4.2 Collective Behavior Analysis . . . . . . . . . . . . . . . . . . . . . . 91 5.4.3 Datasets and Validation . . . . . . . . . . . . . . . . . . . . . . . . . . 94 6 Final Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 References.... .... .... .... ..... .... .... .... .... .... ..... .... 99 Chapter 1 Introduction Enthusiasts from different areas have observed human behavior for many years, decades,orevencenturies.Suchobservationscanproducevaluabledatatobestudied indifferentfields,fromengineeringtopsychology. In the nineteenth century, LeBon (1895) observed that, when part of a group, an individual can abandon her1 own mental identity and assumes the identity of thegroup.Also,theindividualcanhavehis/herjudgmentaffectedasafunctionof collective behaviors around him/her. According to the author, the individuals in a crowdcandiscardtheirownvaluesandeventheirinhibitionsandpresentbehaviors thatshouldnotbeperformediftheywerealone.Suchunusualbehaviorcanengender differentemotionsinpeople.Thesefeelings,forexample,anxiety,jitters,orpanic, make the individual more emotive and, sometimes, irrational. Similarly, Sighele (1954) highlights situations in which people lost reason when in crowds and act againstdifferenttargets,includingtheirownstate.Boththeauthorsdiscussthepower ofacrowd,whichcanpresentanuncontrollableandunpredictableforce. Suchemergent,oftenunpredictable,collectivebehaviorscanoccurwhenpeople arepartofacrowd,andtheycanshareideas,feelings,andthesameorsimilargoals. While the accepted definition of a crowd is a large group of people that are gath- eredorconsideredtogether,recentscientificstudieshaveconsideredcrowdsasan entityabletoself-organize(McPhail1991).Examplesofsuchcollectivebehaviors arethespontaneousformationoflanesofuniformwalkingdirectioninbidirectional flows(Milgrametal.1969),ortheoscillationofthepassingdirectionatnarrowbot- tlenecks(HelbingandMolnár1995).Inaddition,LeBon(1895)claimsthatthereare severalcharacteristicsofcrowdpsychology:“impulsiveness,irritability,incapacity to reason, the absence of judgment of the critical spirit, the exaggeration of senti- ments,andothers.”LeBonclaimed“thatanindividualimmersedforsomelength oftimeinacrowdsoonfindshimself—eitherinconsequenceofmagneticinfluence 1Inthisbookthefeminineforms/her/,/hers/,and/she/willsubstituteformoreawkwardcombina- tionssuchas/him/her/,/his/hers/,/he/she/,and/she/he/. ©SpringerInternationalPublishingAG2017 1 V.J.Cassoletal.,SimulatingCrowdsinEgressScenarios, https://doi.org/10.1007/978-3-319-65202-3_1 2 1 Introduction givenoutbythecrowdorfromsomeothercauseofwhichweareignorant(inaspe- cial state),which much resembles the state of fascination in which the hypnotized individualfindshimselfinthehandsofthehypnotizer.” The definition proposed by Sighele (1954) considers a crowd as a heteroge- neousand inorganicaggregationofpeople.Heterogeneous usuallymeansthata crowdconsistsofindividualsfromallages,gender,anddifferentsocialandcultural realities.Acrowdisconsideredinorganic becauseofitscapabilitytoemergesud- denly without formal control and organization. However, the aggregated structure isimportanttorecognizeinacrowd.AsobservedbyThalmannandMusse(2013), aggregated motion is both beautiful and complex to contemplate: beautiful due to the synchronization, homogeneity, and unity described in this type of motion, and complex because there are many parameters to be handled in order to evoke these characteristics.AccordingtoFruin(1971b,c)crowdbehaviorisaffectedbythespa- tialperceptionofeachindividualconsideringherownknowledgeandintelligence. Whentheenvironmentisknown,theindividualcanmakeadecisionbasedalsoon social and cultural patterns. Personal preferences also affect how close people can gettoeachother.Thespaceoccupiedbypeopleandtheirrelationshipswithothers wasstudiedbyEdwardHall(1966).AccordingtothisAmericananthropologist,who coinedthetermProxemics,eachpersonhasasetofnestedspacesaroundthebody: public,social,personal,andintimate.Thepersonalspacecanvarybasedonthekinds ofinteractionandrelationshipsofthepeopleinvolved.Inaddition,theinterpersonal distanceamongpeopleincrowdscanalsobeaffectedbyindividualcharacteristics suchasgender,ageofeachindividual,andherphysicalstate. We know that some places can be propitious to crowd formation. Such places canincludeairports,trainstations,andpublicareas,forexample.Theexistenceof places propitious for crowd formation, leads governments, managers, researchers, designers, and other professionals to be acutely interested in the development of technologies to improve the security and comfort of those places. One of these relativelynewtechnologiesiscomputationalcrowdsimulation. Crowd Simulation addresses related problems in the entertainment (games and movies) and safety industries. For entertainment, we can easily apply crowd sim- ulation to populate scenes of a game or movie with realistic and dynamic crowds, e.g.,duringsimulatedwarfare,orbyaudienceenhancement toreducetheexpense ofhumanextras.Ontheotherhand,insafetyengineering,avoidanceofinjuryisthe primary concern, and we can observe some open research problems when crowds areconsidered.Inthisbook,weshowthestateoftheartinthisareaandalsopoint outopenproblemsandproposedsolutions.Indeed,ourfocusistodiscussevacuation models,theimportanceofevaluation,andvalidationinthesesituationsandcasestud- ies,e.g.,gamesatstadiums,publicbuildings,trainstations,etc.Suchunderstanding allowsengineerstodesignbetterplacesandalsotofigureoutthebestwaytoguide peoplewhenchoosinganevacuationroute. Variousapproacheshavebeenproposedintheliteratureoverthelastfewdecades motivating the development of different scientific crowd motion models. Such approachesareconcernedwithcomputationalsimulationofindividualmotionbehav- ior, small groups, and crowds. These models were designed based on different 1 Introduction 3 goals and complexity levels. The first known model is a local rules-based sys- tem (Reynolds 1987) able to simulate the behavior of flocks, herds, and schools of animal surrogates called “boids”, where all individuals are homogeneous with respecttotheirruleparameters.Animportantandcurrentaspectinsimulatedcrowd behavior is that individual parameters characterize heterogeneous crowds, where decisionsandactionscanvaryfromonetoanotheranddifferentlyinfluencecrowd evolution.Somerecentpublicationsdiscusstheimportanceofhavingheterogeneous agentsinacrowd(Pelechanoetal.2016).Also,navigationandsteeringbehaviors haveachievedincreasinglyrealisticresults,(Patiletal.2011;Bersethetal.2015a; Boatrightetal.2014). Despitetheserecentdevelopments,therearestillopenopportunitiestoconsider thesimulationprocessforemergencysituationsandapplicationofmethodsinreal life.Thisbookaimstodiscusssomeoftheseaspects,anditisorganizedasfollows. First,wediscusssometheoryandmethodsbehindcrowdsimulation,crowddynamics and evolution in egress situations, and regulations in evacuation processes which can vary from place to place in the world. Then, we focus on crowd simulation technologiesdescribingsomecurrentextantpossibilitiesforpeoplewhowanttowork ordeveloptoolsinthearea.InChap.3wepresentCrowdSim,thesoftwaredeveloped bytheauthors.Wealsodiscussthechallengingproblemofvalidationandcomparison ofsimulationwithreal-lifesituations.InChap.4theCrowdSimsoftwareisthebasis forseveralcasestudyscenariosdrawnfromreal-lifeapplications.Finally,inChap.5 wediscusstheprospectofperformingcrowdmotionanalysisusingcomputervision techniques.