grasshopper add-on - spider web

12
Proceedings of the Ninth International Space Syntax Symposium Edited by Y O Kim, H T Park and K W Seo, Seoul: Sejong University, 2013 SPACE SYNTAX FOR GENERATIVE DESIGN: On the application of a new tool 050 Richard Schaffranek Vienna University of Technologie e-mail : [email protected] Michael Vasku Vienna University of Technologie e-mail : [email protected] Abstract Space Syntax can be used to predict, qualify and quantify aspects of human behaviour of different design proposals and the impact on their neighbourhoods. Therefore some design projects already incorporate Space Syntax during the design process stage (Dursun 2007) by proposing different design alternatives and continually testing and adopting them. In recent years, it has become increasingly common to use physical characteristics like the wind, solar exposure or sound, as driving forces behind a generative design approach. Similar to the various physical models, Space Syntax analysis can be also, some simplifications assumed, processed automatically. Such an approach can be referred to as “inverse design”, design by intents, used as constrains to a form and not as form by itself (Faucher and Nivet 2000). This paper introduces “SpiderWeb” a new tool, under development since 2011, which enables various generative approaches based on Space Syntax. It provides the basics to copy some of the analysis available in “depthmapX”. (“depthmapX” is a multi-platform software to perform a set of spatial network analyses designed to understand social processes within the built environment. “depthmapX” is based on the original “UCL Depthmap” and its developed by Tasos Varoudis at UCL’s Space group. (Space Syntax Limited 2013)) Furthermore “SpiderWeb” extends it to other forms of modelling, applicable for urban and architectural design in three-dimensional space. Keywords: Architecture, Parametric - Generative Design, Grasshopper3D, Space Syntax Theme: Modelling and Methodology Development

Upload: anca-horvath

Post on 16-Aug-2015

288 views

Category:

Documents


15 download

DESCRIPTION

Grasshopper Add-On - Spider Web

TRANSCRIPT

Proceedings of the Ninth International Space Syntax Symposium Edited by Y O Kim, H T Park and K W Seo, Seoul: Sejong University, 2013 SPACE SYNTAX FOR GENERATIVE DESIGN: On the application of a new tool 050 Richard Schaffranek Vienna University of Technologie e-mail : [email protected] Michael Vasku Vienna University of Technologie e-mail : [email protected] Abstract SpaceSyntaxcanbeusedtopredict,qualifyandquantifyaspectsofhumanbehaviourof differentdesignproposalsandtheimpactontheirneighbourhoods.Thereforesomedesign projectsalreadyincorporateSpaceSyntaxduringthedesignprocessstage(Dursun2007)by proposing different design alternatives and continually testing and adopting them. In recent years, it has become increasingly common to use physical characteristics like the wind, solarexposureorsound,asdrivingforcesbehindagenerativedesignapproach.Similartothe variousphysicalmodels,SpaceSyntaxanalysiscanbealso,somesimplificationsassumed, processedautomatically.Suchanapproachcanbereferredtoasinversedesign,designby intents, used as constrains to a form and not as form by itself (Faucher and Nivet 2000). ThispaperintroducesSpiderWebanewtool,underdevelopmentsince2011,whichenables various generative approaches based on Space Syntax. It provides the basics to copy some of the analysisavailableindepthmapX.(depthmapXisa multi-platformsoftwareto performaset ofspatialnetworkanalysesdesignedtounderstandsocialprocesseswithinthebuilt environment. depthmapX is based on the original UCL Depthmap and its developed by Tasos VaroudisatUCLsSpacegroup.(Space SyntaxLimited 2013))FurthermoreSpiderWebextends ittootherformsofmodelling,applicableforurbanandarchitecturaldesignin three-dimensional space. Keywords: Architecture, Parametric - Generative Design, Grasshopper3D, Space Syntax Theme: Modelling and Methodology Development Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 2 Introduction SpiderWeb - the tool introduced in this paper, - enables the use of Space Syntax in parametric modelling and its pursuit in computational design. Parametric design is a technique that enables adesigndependingonconstraints.Constraintsaredefinedassociationsandrestrictions expressedingeometryandwhichhelptofitspecifiedrequirements.Generallyspoken,two typesofconstraints,i.e.geometricanddimensional,canbedifferentiated.Geometric constraints are used to control the relationships of objects in respect to each other. Dimensional constraintsareusedtocontrolthedistance,angle,radiusandlengthvaluesofobjects.A parametricmodelformsthebasisforcomputationaldesign.ComputationalDesignisa procedural,repeatable,mathematicallydefinableprocess.Itisbasedonpreliminarilyspecified quantitativesetsofrulesthatcanbeexploredanddevelopedinaframework.Theframework hastoallowvariationstobegeneratedandevaluatedbasedondefinedandundefined parameters,operations,orrepresentations.Thispaperpresentsdifferentapproachesforusing Space Syntax as a constraint in a computation design process. Motivation Figure1:Background:Rhino5viewport.(bottom-left)graphrepresentationoftheneighbourhoodrelationsbetween differentrooms(walls,rooms,entrance).Thegraphisusedtogeneratefloorplanlayouts(top-right).Ajustifiedgraph (top-left) canbevisualised toanalyse thelayout. Foreground: The Grasshopper3ddefinition, with some SpiderWeb components,tocalculatetheinformationintheviewport.Inthisscreenshotameta-heursiticsolver(Simulated Annealing)isrunningtofindavalidfloorplanaccordingtotheinputgraphwhileminimizingtheperimiterofthe bounding box. TobeabletouseSpaceSyntaxincomputationdesignprocessesitisnecessarytointroduce basicgraph1operationsinsideaframeworkthatallowsmodifyingandgeneratinggeometry parametrically.ForthistheCADsoftwareRhinoceroswaschosen-morepreciselyaplugin namedGrasshopper3D,whichexpandsRhinointhesegmentofparametricandgenerative modelling.2Grasshopper 3D is a visual programming language and therefore quickly to learn and easily to use. In such an environment components are wired together using various inputs to perform e.g. simple/ advanced calculations like the addition of values or the generation of a Voronoi diagrams. 1A graph is a mathematical representation of relations between a set of nodes. If two nodes have a relation they are connected by an edge. An edge has an edge weight. To calculate the shortest path (metric) the edge weight would need to be the Euclidian distance between the nodes connected by the edge. 2RhinocerosandGrasshopper3daredevelopedbyRobertMcNeel&Associates.(www.rhino3d.com; www.grasshopper3d.com)BeyondGrasshopper3dotherenvironmentssuchasDesignScript,currentlyunder development by Autodesk are available. (designscript.ning.com) Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 3 Spiderweb Components SpiderWeb provides the following set of basic calculations/components to create, analyse and perform different operations on graphs (including visual graphs). In the most recent release the components are capable of handling directed graphs as well. Graphs (adjacency representation3) can be generated through different inputs e.g.: GraphFromPointsCreatesagraphfromasetof3dpointsbasedonadistance.Pointsare understoodasnodesofthegraph.Twopointsareconnectedbyanedge,ifwithinaspecified distance. GraphFromLinesCreatesagraphfromasetof3dlines.Thelinesrepresentedges,the endpointsthenodesofthegraph.Ifsettoundirectedtwoedgesareaddedtothegraphfor each line in the set. E1(A, B), E2(B, A) If set to directed one edge is added to the graph. E1(A, B) This allows to model elements like escalators or a one-way street. Graph From Cells Creates a graph from a set of 3d polylines. Each shape represents a node, if two polylines share a boundary the nodes are connected by an edge. GraphFromDatatreeCreatesagraphfromadatatree4.Thepathindicesofthedatatreeare interpreted as an edge list. E.g. this can be used to generate a graph of an axial map. DualGraphCreatesanewgraph,wherenodesrepresenttheedgesoftheoldgraph.Iftwo edges shared a node in the old graph the nodes representing these edges in the new graph are connected by an edge. The cost of the new edge is the angle between the edges or the original graph or if set to topologic 0 for no change in direction and 1 for a turn. In combination with the Shortest Path component this allows to calculate the simplest path. (Turner 2007), (Hillier und Iida 2005) VisualGraphGridCreatesa2dpointgridwithintheboundingrectangleofapolyline. Additionalpolylinescanbeaddedtorepresentobstacles.Alistwithintegervaluesdefinesifa point is inside or outside of an area / an obstacle. Further components allow the computation of different properties of a graph e.g.: BreadthFirstSearch-Preformsabreadth-firstsearch5-algorithmonthegivengraphfroma startingpoint.Thiscanbeusede.g.tocalculatetheintegrationofagivengraph.(Hillierand Hanson 1984, 108) Minimal Spanning Tree Calculates the minimal spanning tree.6 ShortestPathCalculatestheshortestpath(distance)fromarootnodetoallothernodes.If there are multiple solutions (i.e. a regular grid) all possible paths are calculated. 3An adjacency representation describes a graph through its neighbourhoods.This can be done e.g. through a vertex list, where the neighbours of every vertex (node) are stored in a separate index list. 4AdatatreeisaspecialdatastructurewithinGrasshopper3d.Itisbasedonasorteddictionary.Avalueina datatreeisaccessedthroughapath.Thispathisthebasistocreatethegraphe.g.Path:{0;2}willcreateanedge connecting node 0 with node 2; 5Breath-firstsearchisawaytosearchthroughagraphfromagivenrootnode.Inafirststepitsearchesall neighbours of the root node and then continues with the first neighbour found. This is repeated until all accessible nodes, from the root node are visited. 6Atreeisaconnectedgraphwithoutcircles.Aspanningtreeofaconnectedgraphisasubgraph,containingall nodes of a graph without any circles. In a minimal spanning tree, the sum of all edge weights is a minimum. Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 4 RecursiveShadowCasting-Preformsrecursiveshadowcastingonavisualgraphgridand outputs the visual graph. (Turner, Doxa, et al. 2001) Shortest Path Between Points Calculates how often a line segment is part of the shortest path betweeneachpairofgivenstartingpoints;e.g.thiscanbeusedtocalculatethechoiceofa given graph; (Hillier and Iida 2005) Grasshopper3Dalsoprovidescomponentswherescriptscanbewrittenandexecuted.All classes7usedwithinthedescribedcomponentsarealsoaccessiblefromwithinsuchscript components. Structure & Application Thedescribedcomponentsallowcopyingsomemeasurements,providedindepthmapX (Space Syntax Limited 2013), to Grasshopper3d. To do so one needs in depth knowledge about thosemeasurements,sincethecomponentsperformonlythemostbasic graph-based-algorithms.Lessadvanceduserscanapplyclusters8providedtoperformthe differentstandardmeasurementslikebetweenness,closenessorconnectivity.Furthermore, SpiderWebworksinthethirddimension.TherehavebeenotherattemptstoprovideSpace Syntaxmeasurementsinparametricdesignenvironments.TheDecodingSpacesGroup(2012) providessomemeasurementsin2d,aspredefinedcomponentforGrasshopper3d.For generativedesign,however,thisistoorestrictive.Thebasiccomponents,andevenmorethe classesandfunctionsthatcanbeaccessedfromwithinthescriptcomponents,providedby SpiderWeb, on the contrary, allow adapting to various specific tasks: TheEuclidianpositionofthegraphnodesandtheadjacencyrepresentationofthegraphare accessibleasdatastreamswithinGrasshopper3d,unliketheDecodingSpacesplugin. Thereforethegraphcanbecomemorethanthetopologicalrepresentationofa/theformbut canbea/thegenitorofform(seeexamplesbelow:3dCirculationandDigitalAssumption,on circulation and identity). Unlikethewell-definedmeasurementsprovidedbydepthmapX,withSpiderWeb,models canbeadaptedtoincorporatedifferentmeasurementsinmodelsforearlystageplanning decisions.Inthisstagetheexactspatiallayoutisunknown,butthefewconstraintsknowncan helptomakeinformeddecisions(seeexamplebelow:SpaceAllocationPlanningand Communication Design). Differentmodelstodescribeaspectsofhumanbehaviour(e.g.Choice,VGAorfindingsonthe topicofwayfinding(HochmairandFrank2000))canbeincorporatedandaccessedwithinthe samemodeltodescriberelationshipsbetweendifferentlevelsofperception(localglobal, visualperceptionpedestrianflow,spaceallocationpedestrianflow(seeexamplebelow: Labyrinth Runner example). Ifappropriate,simplifiedandautomaticallygeneratedmodelsofspatialarrangementscanbe used, resulting in a loss of precision but a gain in speed. Further only the values needed, can be calculated(seeexamplebelow:LabyrinthRunnerexample).Thisimprovesthespeedofthe calculationandenablestheapplicationofmeta-heuristicsolvers9searchingthroughavast amount of different solutions for specific properties. 7AclassisanabstractmodelofanObject(e.g.aGraphEdge).Aclasscanhavedifferentproperties(e.g.start, endpoint and costs) and methods that can be executed (e.g. inserting a vertex in a GraphEdge, returning two new Edges, or changing the costs of a GraphEdge) 8AclustercanbeunderstoodlikeablockoragroupinCADprograms.ItcombinesasetofGH-componentsintoa single component including inputs and outputs. 9Meta-heuristicsolverscanbeusedtosolveoptimisationproblemswithoutknowingspecificsabouttheproblem. Therefore they can be applied to a variety of problems. Depending on the type, e.g. a genetic algorithm, the use of different strategies e.g. evolutionary principles to search possible solution trying to find the best solution. Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 5 Examples of generative applications Network Improvement Network improvements aim to find small but weighty changes in a given axial, segment or road centrelinerepresentationofanyurbannetworkinordertofitrequirementsdefinedbySpace Syntaxvalues.(Vasku2013)Thereareseveralcasesinwhichsuchanimprovedscenariomight beuseful,likecontrollingandredirectingmovementininformalsettlementsafteridentifying areasofstructuralsegregation.Aredesignmaycausepositiveeffects,aimingtocreatea condition to allow slums to self-correct themselves. (Karimi und Parham 2012) (Karimi, Amir, et al.2007)Thetaskistofindefficientinterventionsthatoppose"spatialdiscrimination"with minimal interventions on build stock. In a first step new connections between existing axial lines are created. In praxis, the consequence of a new street means the removal of existing buildings within the road's new area. Theoretically there are infinite possibilities for new line generations, soonehastolimittheiramounttoamanageableamount.Inthiscasenewconnectionsare drawnuptothemaximallinelengthof350mbyconnectingonly10%ofthemostintegrated axial lines at local radius with the 10% of most integrated axial lines at a global radius. Figure 2:a) initial axial lines map (Karimi und Parham 2012)b) R5integration analysisof initial axial mapc) highlighted new axial lines for improvement d) R5 integration analysis of improved axial line map For every single axial line that is generated, an individual Space Syntax integration analysis at a globalradiusiscalculated.Foreverycalculationdifferentvaluesareextractedandsavedfor later processing, such as: Generation ID (to identify the axial line), Intelligibility (a value defined bySpaceSyntaxtheory),0.25quantileofallintegrationvalues,and0.75quantileofall integration values. Here SpiderWeb in combination with Grasshopper allows the generation of newlines,theirindividualanalysisandfurtheranextractionofselectedvaluesinabatch process. These values are stored in a table to enable their comparison and evaluation. TheIDsofthetoponepercentofeachmeasurement(descendingorder)areselected.The more often an ID is selected the larger the impact of its associated axial line. Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 6 Figure3:Thegraphcomparesintegrationvaluesofthesystembeforeandaftertheintervention.Asnewlineswere generated to improve the network, the total amount of axial lines is increased (x-axis). Figure3comparesvaluesobtainedbyanintegrationanalysisofthesystemwithandwithout the calculated intervention. It clearly shows the significant improvement of integration, which is equallydistributedthroughouttheentirespectrum.Thatimplies,thattheupgradingeffect measured by integration effects most settlers without privileging certain locations. The values of eachaxiallinewithintheinformalsettlementandinitsdirectneighbourhoodaredisplayedin ascending order for their integration value. Labyrinth Runner NotonlymeasurementsproposedbytheSpaceSyntaxcommunitycanbeincludedina generative-parametricmodelwithSpiderWeb,butalsootherassumptionslikewayfinding strategiesinunknownnetworks,asdescribed byHochmairandFrank(2000)canbeutilizedto describeaspectsofhuman behaviourthroughamodel.Suchamodelcanbeused togenerate labyrinths in which one can get lost. For this example three different measurements have been used to define the fitness of a solution, for use with a meta-heuristic solver. Let G(V,E G) be the graph representing the Labyrinth and P = (P0 = A,P1,P2,...,Pn = B; P0..n V) theshortestpath from Ato B.Thecostsc(p)(sumofchangesindirection)foratrip,alongthe shortest path, from A to B must be as high as possible. The more the shortest path p converges to a space filling curve, the higher the costsc(p). The average of Euclidian distances d(p) from all nodes part of the shortest path p from A to B must be as high as possible. This contradicts the way finding strategies of humans described by Hochmair and Frank (2000). In order to create a network-likestructure,themoreedgesn(p),notpartoftheshortestpathfromAtoB,the better.Thelargerthefitnessf2=c(p)*d(p)*n(p)is,thehardertofindthepaththroughthe labyrinth.Thegeneratedlabyrinthswereusedaslevelsforacasualgame.10Theplayerpaths were traced in order to analyse and verify the generated spatial layout. The traces showed that if in an unknown network a decision to change direction is only made if there are no alternatives. Someplayersplayedthesamelevelmoreofteninarowimprovingtheirunderstandingofthe spatial layout constantly. But when they replayed the level, after a few minutes, they got lost again. 10The game can be played at http://wp10478398.server-he.de/LabyrinthRunner. To get access to the collected data please contact the author. Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 7 Figure4:ShowingmultiplerunsbyaplayernicknamedLeDove.Thesixminutesinbetweenthelasttwoplaysofthe LevelRoomToRoom6x6LeDovetriedtosolveotherlevels.Whenhereplayedthislevelaftersixminuteshepartly lost orientation. Furtherthedatacollectedmadeitpossibletoproposeasimpleproceduralalgorithmto generatelabyrinths.Thissimplealgorithmcouldalsobeusedtogeneratelayoutswithmore than two endpoints or within different boundaries. (Schaffranek 2013) Digital Assumption, on Circulation and Identity ThisexamplecontinuestoexplorethepossibilitiesoutlinedintheLabyrinthRunnerexample using different fitness functions. It was developed together with students within a studio project. The students where provided with a parameterised-line-based -model of a building. The task for the students was to use the model and define an algorithm that would generate geometry from the line system, which should help to orientate and define the identity of the different zones of circulation.Togeneratethesedifferences,numericaloutputsofdifferentmeasurementswere used. In the parameterised-line-based model, each line would represent a certain area of the building (e.g.asquareofabout4by6meters).Itcouldbesettooneoffourstates:0-lineisdeleted, 1-lineisflat,2-lineisgoingup,3-lineisgoingdown;Utilizingameta-heuristicsolver,different properties(fitnessfunctions)couldbesearchedfor,withinthissimplifiedlinesysteme.g. shortestpathbetweenallpoints,largestdistancebetweenlocalcentres,highestdifference between the choice value of the most used line and the average choice value. Proceedings of the Ninth International Space Syntax Symposium, Seoul, 2013 R Schaffranek and M Vasku: Space syntax for generative design050: 8 Figure5: Example with the largestdistance between local centers (pink circless); A doubleloop emergedas circulation, generating two areas which are closer to all other points than their neighbours; white dotted lines:underlying graph for calculating the different properties (e.g. avg distance between all points); blue: smallest avg. distance, red: heighest avg. distance to all points; The model of the building would be considered inadequate for an accurate analysis, but through simplificationitwaspossibletocalculatethefitnessinanacceptabletime(