29
Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 Cheng Zhang

Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

  • Upload
    others

  • View
    18

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Crowd Simulation, Motion Planning

CSE888.14X – Autumn 2008

Cheng Zhang

Page 2: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Mo#va#on(Crowd)

•  Entertainment(moviesandgames)•  Architecture•  Training•  Security•  Sociology(crowdbehavior)•  Physics(crowddynamics)

Page 3: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Approaches(Crowd)

•  Par#clesystems•  Agentbasedmodels

•  Probabilitynetworks•  Socialforcenetworks•  Flowandnetworkmodel

•  ……

Page 4: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Approaches(Cont.)

•  Par#clesystems:thecharactersareaFachedtopar#clepointsthatarethenanimatedbyvectorfields.– Pros:inexpensive,canbedoneinmost3DsoLwarepackages.

– Cons:limita#onofthebehaviors notveryrealis#c

Page 5: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Approaches(cont.)

•  Agentbasedmodels:

•  Twosurvey‐papers:–  THALMANND.,O’SULLIVANC.,CIECHOMSKIP.,DOBBYNS.,Popula'ng

VirtualEnvironmentswithCrowds.Eurographics2006TutorialNotes.

–  SCHRECKENBERGM.,SHARMAS.D.,PedestrianandEvacua'onDynamics.Springer,2001.

•  Pros:Thissystemismuchmorerealis#cthanpar#cleSystems.

•  Cons:expensivecomparedtotheprevious.

Page 6: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Key Issues

•  Large-scale crowd in real time

•  Richness of behaviors in a crowd

•  control

•  computational cost

Page 7: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Papers(crowd)•  REYNOLDS C. 1987, Flocks, herds, and schools: A distributed behavior model. In Proceedings of ACM

SIGGRAPH 87 (July 1987), Annual Conference Series, ACM SIGGRAPH.

•  REYNOLDS, C. 1999. Steering behaviors for autonomous characters. In Proceedings of Game Developers Conference 1999,763–782.

•  MUSSE, S., AND THALMANN, D. 2001. Hierarchical model for real time simulation of virtual human crowds. IEEE Transactions on Visualization and Computer Graphics 7, 2, 152–164

•  ULICNY, B., CIECHOMSKI, O., AND THALMANN, D. 2004. Crowdbrush: Interactive authoring of real-time crowd scenes. In Proceedings of the 2004 ACM SIGGRAPH/Eurographics symposium on Computer animation, 243–252.

•  Anderson, M. McDaniel, E. and Chenney, S. 2003, Constrained animation of Flocks. SCA

•  SUNG, M., GLEICHER, M., AND CHENNEY, S. 2004. Scalable behaviors for crowd simulation. Computer Graphics Forum 23, 3, 519–528.

•  Mankyu Sung and Lucas Kovar and Michael Gleicher, 2005, Fast and accurate goal-directed motion synthesis for crowds, Proceedings of the 2005 ACM SIGGRAPH/Eurographics symposium on Computer animation

•  Adriano Rinaldi, Alessio Malizia, Rick Parent, 2006, Modeling Behavior in a School of Fish by Fast Synthetic Distributed Vision

•  Adrien Treuille, Seth Cooper, Zoran Popovic , 2006, Continuum Crowds, SIGGRAPH 2006

•  R. McDonnell, S. Dobbyn and C. O’Sullivan, 2006, Crowd Creation Pipeline for Games

•  Wei Shao and Demetri Terzopoulos, 2007, Autonomous pedestrians, Graph. Models, vol. 69.

•  Xiaogang Jin, X., Wang, C., Huang, S., Xu, J. 2007, Interactive Control of Real-time Crowd Navigation in Virtual Environment, VRST '07: Proceedings of the 2007 ACM symposium on Virtual reality software and technology

•  N. Pelechano, J.M. Allbeckand N.I. Badler, 2007, Controlling Individual Agents in High-Density Crowd Simulation

Page 8: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Composite Agents, SCA 08Yeh, H., Curtis, S., Patil, S., Berg, J. Manocha, D., and Lin, M.

•  Acompositeagentconsistsofabasicagentthatisassociatedwithoneormoreproxyagents.

•  Theideaistoinjectintangiblefactorsintothesimula#onbyembodyingthemin"physical"formandrelyingonthesimulator’spre‐exis#ngfunc#onalityforlocalcollisionavoidance.

Page 9: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Composite Agents (cont.)Yeh, H., Curtis, S., Patil, S., Berg, J. Manocha, D., and Lin, M.

•  Algorithm–agentbased•  Naviga#on–globalroadmap

•  Collisionavoidance–reciprocalvelocityobstacles

Page 10: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Composite Agents, SCA 08

•  Compositeagentformula#on

Page 11: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Composite Agents (cont.)

Page 12: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Composite Agents (cont.)

•  Limita#on:– behaviorscomplicatedcommunica#onorgroupcoordina#on.

– compositeagentsrelyonthemechanismprovidedbytheunderlyingplanningsystem(e.g.collisionavoidance),thislevelofindirec#ondisallowsprecisecontrolovertheexactnatureoftheagentinterac#ons.

Page 13: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Group Motion Editing SIGGRAPH08

Taesoo Kwon, Kang Hoon Lee, Jehee Lee, Shigeo Takahashi

•  Areal‐#meapproachtoedi#nggroupmo#onasawholewhilemaintainingitsneighborhoodforma#onandindividualmovingtrajectoriesintheoriginalanima#onasmuchaspossible.

Page 14: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Group Motion Editing (cont.)

•  Thegraphstructure‐ver#cesrepresentposi#onsofindividualsatspecificframesandedgesencodeneighborhoodforma#onsandmovingtrajectories.

•  Edi#ngopera#ons– Deformgroupmo#onbypinningordraggingindividuals.

– S#tchormergemul#‐groupmo#onstoformalongerorlargergroupmo#onwhileavoidingcollisions.

Page 15: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Group Motion Editing (cont.)

Page 16: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Group Motion Editing (cont.)

•  Advantage‐ausercaninterac#velymanipulatemul#plecharactermo#onsasawholeands#llhavedirect,precisecontroloverindividualtrajectories.

•  Limita#on‐Alargedeforma#onofagroupmo#oncanleadtounnaturalspeedup/slowdownofindividualmo#ons.

Page 17: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Clone Attack! Perception of Crowd Variety SIGGRAPH 08

McDonnell, R., Larkin, M., Dobbyn, S., Collins, S.,O'Sullivan, C.

•  Study - Perception of variety in crowd.

•  Issue - when simulating large crowds, the models and motions of many virtual characters are often cloned. However, What is the proper degree of the duplication in terms of the perception of variety in crowds?

Page 18: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Clone Attack! (cont.)

•  BaselineExperiments(singleclone)–studythefactorsthataffectthepeople’sabilitytoiden#fyasinglepairofclonesincrowd.

•  Mul#pleCloneExperiments– Mainhypothesisisthatincreasingthenumberofclonesofasinglemodelormo#onwillmakeclonepairseasiertofind.

Page 19: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Clone Attack! (cont.)

Page 20: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Clone Attack! (cont.)

Page 21: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Clone Attack! (cont.)

Summaryofeffects:•Appearancecloneswereeasiertodetectthanmo#onclones•Increasingclonemul#plicityreducedvarietysignificantly•Noappearancemodelwasmoreeasilydetectedthanothers•Certaingaitsweremoredis#nc#vethanothers•Colormodula#onandspa#alsepara#oneffec#velymasked

appearanceclones•Combinedappearance/mo#oncloneswereonlyhardertofindthan

sta#cappearancecloneswhentheirclonedmo#onswereout‐of‐step

•Appearancecloneswerealsohardertofindwhencombinedwithrandommo#ons

•Mo#oncloneswerenotaffectedatallbyappearance,evenwithrandomappearances

Page 22: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Mo#onPlanning

•  Ahardprobleminadynamicenvironment•  Pathplanning‐akinema#cprobleminvolvingthecomputa#onofacollision‐freepathfromstarttogoal.

•  Velocityplanning‐theconsidera#onofrobotdynamicsandactuatorconstraints.

Page 23: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Approaches(planning)

•  Velocityobstacle‐P.FioriniandZ.Shiller,“Mo#onplanningindynamicenvironmentsusingvelocityobstacles,”Int.JournalofRobo#csResearch,vol.17,no.7,pp.760–772,1998.

•  Reciprocalvelocityobstacle‐VanDenBergJ.,LinM.,Manocha,D.,“Reciprocalvelocityobstaclesforreal#memul#‐agentnaviga#on.Proc.ofIEEEConferenceonRobo?csandAutoma?on(2008).

Page 24: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Mo#onplanningindynamicenvironmentsusingvelocityobstacles

•  AistherobotandBisobstacle.•  Eachhavearespec#vevelocity,

VAandVB.•  Assump#on:Instantaneous

state(posi#onandvelocity)iseitherknownormeasurable.

•  Assump#on:Anobstaclesmaintainsaconstantvelocityatleastforagiven#me.

Page 25: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Motion planning in dynamic environments using velocity obstacles

Page 26: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Motion planning in dynamic environments using velocity obstacles

•  CollisionCone:setofcollidingrela#veveloci#esbetweenA’andB’.

•  V(A,B)istherela#vevelocityofA’withrespecttoB’,andλ(A,B)isthelineofV(A,B)

•  Theveloci#esthatliebetweenλrandλfwillcauseacollisionbetweenA’andB’.

•  Intui#onbehindvelocityobstacles:Theconerepresentsthosesetofveloci#esthatwillcausetherobottocollidewiththeobstacle.

Page 27: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Reciprocalvelocityobstaclesforreal#memul#‐agentnaviga#on

•  ReciprocalVelocityObstacle‐implicitlyassumesthattheotheragentsmakeasimilarcollision‐avoidancereasoning.

•  Underthisassump#on,theproposedframeworkisguaranteedtogeneratesafeandoscilla#on‐freemo#ons.

Page 28: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Reciprocalvelocityobstaclesforreal#memul#‐agentnaviga#on

•  ReciprocalVelocityObstacle‐insteadofchoosinganewvelocityforeachagentthatisoutsidetheotheragent’svelocityobstacle,wechooseanewvelocitythatistheaverageofitscurrentvelocityandavelocitythatliesoutsidetheotheragent’svelocityobstacle.

Page 29: Crowd Simulation, Motion Planningweb.cse.ohio-state.edu/~parent.1/classes/888/Au08/Cheng.pdf · Crowd Simulation, Motion Planning CSE888.14X – Autumn 2008 ... • Sociology (crowd

Futureworks

•  Real‐#mecrowdsimula#onandplanning•  Moredynamiccontrol

•  Morevarietyofbehaviors

•  Trade‐offbetweenautonomyandtheanimatorcontrols