DEVELOPING PARAMETRIC BUILDING MODELS – THE GANDIS USE CASE

International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXVIII-5/W16, 2011 ISPRS Trento 2011 Workshop, 2-4 March 2011, Trento, Italy DEVELOPING PARAMETRIC BUILDING MODELS – THE GANDIS USE CASE Wolfgang Thallera , Ulrich Krispela , Sven Havemanna , Ivan Redi c , Andrea Redi c , Dieter W. Fellnera,b a Institute of Computer Graphics and Knowledge Visualization (CGV) Graz University of Technology, Inffeldgasse 16c/II, 8010 Graz, Austria (w.thaller, u.krispel, s.havemann, d.fellner)@cgv.tugraz.at http://www.cgv.tugraz.at/ b Fraunhofer IGD & Darmstadt University of Technology, Fraunhoferstrasse 5, 64283 Darmstadt, Germany c http://www.ortlos.org/ ORTLOS Space Engineering, Gleisdorfergasse 6, 8010 Graz, Austria. KEY WORDS: Geometry, Parameters, Graphics, Visualization, Software ABSTRACT: In the course of a project related to green building design, we have created a group of eight parametric building models that can be manipulated interactively with respect to dimensions, number of floors, and a few other parameters. We report on the commonalities and differences between the models and the abstractions that we were able to identify. 1 INTRODUCTION The creation of building models is a tedious task. Sophisticated architectural software such as AutoCAD (Autodesk), ArchiCAD (Graphisoft) or Revit (Autodesk) provides architects with various tools for creating detailed plans and sections for the construction process. This software is typically used after the initial form finding process, i.e., when the building concept is clear, and function, stability, and style are defined (or, following Vitruvius, utilitas, firmitas, venustas). For the early design process all sorts of rapid modeling tools are used, most notably Maya or 3D Studio Max (Autodesk), or simple tools like SketchUp (Google), or Rhino (McNeel) for freeform architecture. In this paper we present results from the GANDIS project, which attempts to close the gap between early design and constructionready planning in a different way, namely using procedural building templates. An important drawback of the conventional design process is that some implications of the initial design become apparent only after construction planning. The energy footprint of a building, for instance, depends on the A/V ratio (surface to volume), the glass proportion, the number of floors, but also of the wall insulation standard. So the requirement was to create an interactive planning tool that allows for computing characteristic key values from a set of prototypical parametric buildings. Even if the prototype buildings are only roughly similar to the building to be planned, the tool allows judging the impact of changes in the building on the energy footprint. This allows answering questions such as: How compact should the building be, what happens if we create this overhang, and how expensive is it to compensate energy loss by better insulation on that wall, etc. We consider parametric building templates to be useful for many applications other than energy design. The examples created in this case study are simple, but once the approach is clear, more targeted and elaborate building models can be created using the same approach. The objective of this paper is to illustrate the process of developing parametric building models, in the hope that others can learn from it. We seek to provide an enabling technology for interactively changeable complex procedural building templates. The process starts from a set of example buildings provided by architects (Fig. 1). They span the design space that is to be parameterized. To define this space unambiguously, however, required many discussions, e.g., on the parameter minima and maxima, or to clarify which special cases should be prevented to guarantee that the building created remains valid. The next step was to develop first simple buildings, and then to successively extend them to create more complex ones. This inductive process is marked by continuous refactoring in order to find a set of re-usable parametric sub-constructions (doors, windows, floor plan processing). So our experience was that developing the first few models took most of the time, while later the toolset and thus, the design space, were powerful enough to create more elaborate models faster. The Office series of templates was created from the Residential series merely by replacing the facade decoration function (see sec. 5.6). 2 RELATED WORK If the creation of a valid 3D building model is tedious, then even more so is the creation of a parameterized building. It requires a description of the construction process from input parameters to building elements. Which technologies can be used for this? First of all, almost every 3D modeling software package includes a scripting language to automate construction processes, using languages like C# (Revit), AutoLisp (AutoCAD), MEL (Maya), all from (Autodesk, 2011), Ruby for SketchUp (Google, 2011), 163 International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXVIII-5/W16, 2011 ISPRS Trento 2011 Workshop, 2-4 March 2011, Trento, Italy (a) (b) (c) (d) Figure 2: In the GANDIS project buildings are grouped into sets with similar shape. The following basic shapes are used: rectangular 2(a), L-shaped 2(b), atrium 2(c) and a freeform 2(d) type. (a) rectangular Figure 1: Exploring the design space of the “residential” buildings: static reference models created with conventional modeling software (Google SketchUp) served as guideline for abstraction. or RhinoScript (Robert McNeel & Associates, 2011) etc. We also use a programming language (GML), but it was primarily designed as an efficient shape description language, and not as an add-on to a graphical 3D editor. Grasshopper, a Rhino plugin, is a graphical dataflow editor that allows designers to explore algorithmic shapes without programming. Compared with programming, this approach limits the expressiveness with respect to possible abstractions. In our approach, to find good abstractions is the goal. Only very few approaches can be found in the scientific literature. Chevrier et al. use parametric models to reduce the time for reconstructing heritage monuments (Chevrier et al., 2010). They created a Maya extension for parametric architectural elements. Users can select from a range of predefined parametric models, and after giving an initial estimate, the parameters can be adjusted using measurements from point clouds. Finkenzeller (Finkenzeller, 2008) separates the coarse structure of a building from its appearance, which can be chosen from a given set of styles. The method uses annotated floor plans to define the coarse outline. All information is stored in a graph structure, together with basic architectural information, the locations of walls and balconies etc. The system can generate high-quality models with an appealing level of detai (...truncated)