Architectural lighting ambience intentions for inverse rendering Vincent Tourre, Jean-Yves Martin and Gérard Hégron Laboratoire CERMA (UMR CNRS 1563), École Nationale Supérieure d’Architecture de Nantes, [vincent.tourre|gerard.hegron]@cerma.archi.fr, [email protected]

Abstract : This paper presents a research in progress dedicated to the development of a Computer Aided Architectural Design tool (CAAD) which integrates design by ambience intention concept. To deal with digital design by ambience intention, we examine intention expression of daylighting ambience. In an attempt to add lighting quality to expression features, lighting contrast and direction descriptors are introduced. These descriptors are included in a tool so as to express architectural daylighting ambience intentions through scene lighting properties. As part of a computer aided opening design framework, this tool supply a set of constraints intended to be used in an inverse lighting model. Keywords : Design by intention, Daylighting ambiance, Intention expression, CAAD tools, Inverse rendering.

1 Introduction Architectural and urban ambience concept describes human perception of a built environment through physical phenomenons. Design by ambience intention can be seen as the expression of ambience intentions in the early stages of conceptual design, and the use of these expressions as guidelines to design the building. In current CAAD tools, lighting simulation from a geometric model allows to visualize realistic lighting ambience. However, designing by ambience intention with these tools reveals

a major drawback: a completely known architectural project is required to compute lighting simulation, although this project is precisely being designed from lighting ambience intentions. The integration of design by ambiance intention in CAAD tools leads to digital design by ambience intention concept. However, such an approach needs an lighting ambience intention translation in scene lighting properties, and a lighting description processing to obtain useful results for architects. We intend to give to architects an expression tool in order to reveal their daylighting

ambience intentions, with a view to computing expressive geometrical guidelines for the design. Section 1 presents intention expression features in previous works. Section 2 delineates our approach to materialize lighting ambience intentions in geometric scene properties. The section 3 shows an expression tool for lighting ambience intentions, which combines daylight ambience, intention sketching, and photometric data. In section 4, we present further work, and discuss the use of an inverse lighting technique to compute raw geometrical solutions.

2 Previous works This section presents some description methods of lighting ambience intentions in inverse rendering techniques. Lighting description has a real impact on architect’s intention expression and input data of inverse lighting model. These approaches can be classified following lighting data criterion, which describes lighting properties displayed. These Approaches could also be classified depending on problem type[6], according to whether the required element depends on the scene geometry - inverse geometry, the reflectance properties of surfaces - inverse reflectance, or the position and intensity of light sources - inverse lighting.

2.1 Direct daylighting Architects use direct sunlight to cast shadows and to show sun lighted areas following date and hour. The intention expression of direct light in previous works [8][9][10] has been done by the delimitation of light presence area,

suggesting that other scene parts are not directly lighted. A lighting ambience is a combination of direct light from the source and indirect light from reflections. Direct lighting properties and proportion of direct and indirect lighting have a great influence on the lighting ambience quality. This proportion has been obviously handled in inverse reflectance problem [12], as it determines surface emittance. But neither inverse reflectance intention expression nor previous works on inverse lighting were designed to express these lighting properties. The problem lies in clearly distinguishing direct from indirect light in lighting ambience. Daylighting sources [5][10] are dynamics, as they depend on climate, geographical position, date and hour. Moreover, daylighting sources are not configurable, as we cannot control their position and intensity. So the problem is not an inverse lighting problem, but an inverse geometry or inverse reflectance problem.

2.2 Quantities and quality Measurable lighting data (intensity, emittance, luminance, etc.) describe light exchanged between surfaces or viewed by users. Their expression means have been based on explicit lux amount [1], declarative modelling [3], implicit light sketching [4][9][10] or pictures [6][12]. Costa’s scripts [1] allowed to specify light and shadow, and lighting direction, to model visual comfort. Moeck’s work on object perception [7] handled direction data through object’s appearance. The direction gives an information very useful, as intensity or luminance, but is not present in graphic interfaces.

Figure 1: A design by ambience intention framework : Intention expression is the first stage of the process, and condition input data in inverse lighting.

Lighting quality was explicitly mentioned in few works [5], may be due to the difficulty to deal with subjective impressions. Besides it is difficult to measure a quality, it can be expressed through a set of quantities. But the same quality can be represented by very different properties according to spatial configuration.

3 A design by ambience intention framework 3.1 Setting the problem At CERMA laboratory, research about design by intention is a long term project, developed by Siret[10] for sun lighting, Nivet[2] for visual accessibility, and Houpert[11] for integration of this method in current CAAD tools. Our goal is to provide an aid to geometry design in early stages of conceptual design, by the translation of lighting or visibility ambience intentions in geometric scene properties.

This goal leads us to add new features to the project framework: intention sketching and influence of skylight and built environment. Indeed, natural lighting participates in the relationship between building and environment. This relationship depends on geographical localisation and climate, and is defined by the architect during the early stages of conceptual design. We want to point out that the architect can express more than light levels through lighting ambience intentions. This problem can be seen as the gathering of three elements: daylighting effects, inverse lighting and early stage of architectural design. Daylight source is an extended, heterogeneous and dynamic source, which has an influence on indoor lighting ambience as well as surrounding lighting. External built environment is supposed to be known, and daylight reflections coming from this environment can be considered as secondary light sources. The question of inverse geometry from sunning intentions has been handled by Siret[10], so we focus on difficulties raised by diffuse

lighting from skylight. The lighting ambience provided by skylight can be described in space and time dimensions, but lighting distribution on surfaces cannot be a binary expression like in sun lighted areas descriptions. Inverse lighting method needs to be adapted to match with architectural design constraints. Architects do not ask for an "optimal" solution resulting from an optimisation technique, and needs more guidelines during the design. In inverse rendering framework, this is an inverse geometry problem as we are looking for scene geometry. But this geometry defines also source visibility, and thus it can be seen as source positioning and proportioning problem.

3.2 Framework description Our model (Fig.1) is intended to be used in early stages of conceptual design, when a representation of architectural ambience intentions and a spatial layout are known. We choose to deal with natural light coming from the outside into the designed space, which encompass skylight and reflections from external built environment. The initial direction of diffuse daylight has a great influence on lighting ambience. This lighting direction is mainly conditioned by opening properties (orientation, position, size). We thus focus our research on the lighting direction of lighting ambience, and the computer aided design of openings. The first step of design by ambience intention process is to provide a lighting ambience intention description. The objective is to allow the lighting quality expression as lighting properties with a graphical interface. We propose a tool that allow lighting ambience intention description in a given spatial layout, with

a spatial distribution of daylight on a surface, precision of lighting direction and indication of contrast between surfaces. The second step is an inverse lighting simulation, that allows to transform light properties of an architectural space in geometry and materials properties of this space. We consider openings as intermediary sources with heterogeneous, anisotropic light distribution. Therefore, this is an inverse lighting problem where source position and size actually define space geometry. We can use source positioning methods in inverse lighting framework to compute geometric solutions. Then, the presentation of these raw geometric solutions due to a first interpretation allows the designer to continue his own research of architectural solutions.

4 Expression of ambience intentions We expose in this section a tool to express lighting ambience intentions through spatial lighting properties. We discuss which data have to be present, how they are represented, and how the architect can interact with them.

4.1 Lighting data As we said in previous section, we deal with natural light coming into the designed space. This light comes from skylight, the main light source, and reflections of external built environment, considered as a secondary light source (Fig.4). The direct lighting from these sources is defined only by opening properties. A description of this lighting allows to work directly on opening design.

Figure 2: A direction descriptor is placed on the back face of designed space (Origin point). The intersection between descriptor’s bounds and opening surface delineate a constraint.

The light received on designed space surfaces (e.g., floor, working desk, virtual wall) is described by an heterogeneous spatial distribution. This distribution represents luminance, an absolute light quantity in Lux, and is specified over a time period. Luminance minimal and maximal values can be specified to express light and shadow intentions, respectively. We define the contrast as a lighting proportion, which is a relation between areas. This relation creates dependencies between several areas, on a single surface or different surfaces. A lighting direction is described with a main direction, a solid angle and a luminance value. The solid angle shape is a cone, to give a rough direction indication, or a set of boundary directions, that yet defines an opening form (Fig.2). This shape is centred on the main direction which indicates a source position relative to the surface. Lighting quality is generally closely related to perceived light, and not to direct light. As we aim to aid space design in relation with daylighting, the concept of quality is viewed

as the interrelation between designed space and natural lighting. The preceding relationship is expressed through the concept of direct lighting quality. We argue this quality is better understood with a light decomposition in direct and indirect lighting, to focus on direct lighting.

4.2 Lighting descriptors The objective is to respect lighting ambience intentions, according to their values, directions or period. An area can be described in terms of absolute (luminance) or relative (contrast) lighting quantities. It is possible to set the luminance interval of an area by relating opening properties to direct daylighting properties. Then, it is possible to set contrast interval by relating luminance intervals together. This problem can be seen as a constraint satisfaction problem. A luminance constraint contains a value descriptor (minimal and maximal), a direction descriptor (main direction and bounds) and a time descriptor. This constraint influences

Figure 3: A direction descriptor is reduced by surrounding in order show sky visibility. The designer have visual clues of direction bounds so as to choose lighting intention direction.

opening properties on a specific area of designed space. This area is defined by the intersection between solid angle delimited by direction bounds and designed space. Area opacity is set following value descriptor. The permanency of all these opening properties is defined by time descriptor. A contrast constraint contains area relation descriptors. These descriptors define a relationship between several areas, supposing that at least one of these area is constrained by a luminance constraint. Descriptor domains are reduced by the situation (geographic position, surroundings, etc.), then by designer’s intentions through graphic interface (Fig.3). A luminance or contrast constraint graph can be produced to visualize the system. The resulting set of constraints is intended to be used as input data of an inverse lighting model. This model is based on inverse direct lighting solution proposed by Jolivet[3].

4.3 Graphic interface Light and form are perceived simultaneously, and graphic interface allows to represent and distinguish them. Indeed, direct lighting is not well understood with realistic rendering, because it is not directly perceived, and it is difficult to abstract reflections. Thus, a photometric rendering is used to supply a more precise representation (Fig.4). This photometric rendering allows to represent all lighting data present in previous paragraphs. Combined appearances of these data can be seen as a representation of direct lighting quality. Luminance minimal and maximal values are represented with a copy of described area raised above this area. Contrast is represented by the spatial light distribution on a single surface in order to show which surface areas are proportionally more or less lighted (Fig.4). On different surfaces, contrast is displayed with an indicator sliding on a line between surface centres.

Figure 4: Left Source definition : sky and built environment are set around the outline of project (blue

box in the centre). Right Surface luminance (red and yellow) and contrast (grey level) imported from textures.

Lighting direction is a symbolic data and needs some representation conventions. The representation of one direction with the associated solid angle is indeed not a problem. But the representation of several directions can be a bit confusing, in particular when these directions do not start from the same surface. We propose to show only the intersection of solid angles with designed space to clarify multiple lighting direction display. We keep the idea of sketching lighting intentions directly in virtual scene as in Schoeneman[9]. The sketches describe direct lighting on surface, and are done on orthographic or perspective view. The architect can also imports a texture (picture or sketch) that can be assigned to surfaces. This interface allows architects to sketch their lighting ambience intentions on a partially known environment.

tools through using computer graphic techniques. Our main contribution concern the way of sketching daylighting ambience intentions with contrast, lighting directions and an approach of direct lighting quality. Our choices are justified by the lack of such work in previous works: an alliance of graphic interface, photometric data, and symbolic data. These ambience intentions are intended to be used as input data for an inverse light simulation model. This research currently goes ahead with the adaptation of an inverse lighting model to match architectural design needs, and the computation of missing architectural environment part.

We intend to use a constraint based simulation model that would provide a set of geometrical raw solutions. Therefore we have to show several interpretable geometric configurations resulting from the preceding computa5 Further works tion. The presentation of these solutions will This research links a cultural heritage method, bring us to an information visualization probdesign by ambience intention, and CAAD lem.

References

Painting with light. In Proceedings of the 20th annual ACM Conference on Computer graphics and interactive Techniques (SIGGRAPH), pages 143–146. ACM Press, 1993.

[1] A. C. Costa, A. A. Sousa, and F. N. Ferreira. Lighting design : A goal based approach using optimization. In Rendering Techniques, [10] Daniel Siret. A generative computer tool pages 317–328, 1999. to model shadings and openings that achieve [2] Didier Faucher and Marie-Laure Nivet. Playsunlighting properties in architectural design. ing with design intents : Integrating physiIn Development and application of Computer cal and urban constrants in cad. In ACADIA techniques to environmental studies VI (Pro98 proceedings : Do Computers Make a Difceedings of ENVIROsoft’96), pages 695– ference in Design Studio?, pages 118–137, 704. Computational mechanics Publications, 1998. 1996.

[3] Vincent Jolivet, Dimitri Plemenos, and [11] Houpert Sylvain. Svr 1, a software to Patrick Poulingeas. Inverse direct lightsolve solar constraints: A help to better take ing with a monte carlo method and declarainto account architectural, environmental and tive modelling. In International Conference ecological problems. In PLEA 2002 : design on Computational Science (2), pages 3–12, with the environment, pages 170–175. James 2002. and James Publisher, 2002. [4] J. K. Kawaï, J. S. Painter, and M. F. Cohen. [12] Yizhou Yu, Paul Debevec, Jitendra Malik, and Tim Hawkins. Inverse global illuminaRadioptimization - goal based rendering. In tion: Recovering reflectance models of real Proc. of SIGGRAPH-93: Computer Graphscenes from photographs from. In Alyn ics, pages 147–154, Anaheim, CA, 1993. Rockwood, editor, Siggraph99, Annual Con[5] Ardeshir Mahdavi and Liana Berberidouference Series, pages 215–224, Los Angeles, Kallikova. A generative simulation tool for 1999. Addison Wesley Longman. architectural lighting. In 4th International Conference on Building Simulation, 1995. [6] Stephen R. Marschner. Inverse rendering in computer graphics. PhD thesis, Cornell University, 1998. [7] Martin Moeck. Constraint satisfaction software for architectural lighting design. Leukos, 1(1):101–127, 2004. [8] Pierre Poulin and Alain Fournier. Lights from highlights and shadows. In Proceedings of the 1992 symposium on Interactive 3D graphics, pages 31–38, 1992. [9] Chris Schoeneman, Julie Dorsey, Brian Smith, James Arvo, and Donald Grennberg.