Light Pen

Sketch input in 3D to a knowledge based lighting systemSketching light in 3D

Paper ID: F_999Thomas Jung, Mark D. Gross, Ellen Yi-Luen. Do

NO Affiliation is required for full paper submissionDesign Machine Group, University of Washington, Seattle

Keywords: Lighting design, knowledge-based system, sketching, 3D virtual environment

Abstract: We describe a lighting design system based ondriven by sketching on 3D virtual models. Conventional lighting design tools simulate the lighting effects of design decisions such as window locations, surface treatments, and fixture placement. Light Pen takes the inverse approach by allowing the designer to indicate desired illumination on a 3D model. This Conventional lighting design tools simulate the lighting effects of design decisions such as window locations, surface treatments, and fixture placement. Light Pen takes the inverse approach by allowing the designer to indicate desired illumination. This serves as input to a knowledge-based lighting design system, which recommends solutions on what light to use and where to place it. Light Pen does not attempt to produce an accurate depiction of the space to be lit; rather it proposes lighting design options based on the designer's expressed intentions and on the geometry ofLight Pen the space. serves as input to a knowledge-based lighting design system, which recommends what lights to use and where to place them,. Light Pen does not attempt to produce an accurate depiction of the space to be lit; rather it proposes lighting design options based on the designer's expressed intentions and on the geometry of the space.

1 MOTIVATIONS

Architectural lighting design is an important area of expertise that has implications for health, safety, and sustainable energy use. Until recently, lighting design has been viewed as a specialty area in building design but over the last twenty years—and especially since the California energy crisis of 2001—lighting design has gained prominence as, increasingly, codes and guidelines for public buildings push for more economical, ergonomic, and efficient lighting systems. Much has been accomplished by adopting more efficient lamps (such as compact fluorescents) and switching systems and by recognizing the potential for using daylighting in place of electric lighting. In addition to these technical solutions, lighting design itself can help: by placing illumination where it is needed rather than simply illuminating an entire space at work-light levels. Here is where lighting design expertise comes into play, and hence, intelligent support for lighting in computer aided design.

Computer support for architectural lighting design has been improving steadily since the early days of computer graphics [Greenberg 19761974, Blinn 1976, Newell 1977], growing out of energy simulation programs such as DOE-2 and photorealistic rendering techniques that enable a designer to propose and then simulate design configurations. Work on improving the simulation of lighting effects using, for example, ray tracing and radiosity techniques, has progressed, to the point where lighting design applications (e.g., Lightscape, EcoLumensLumen Micro) are used in professional lighting design alongside traditional physical simulations. Although simulation applications systems help lighting designers understand the consequences of proposed configurations, they they also require the lighting designer to specify a complete configuration of lamps and identify the colors, surface treatments, etc. of the architectural geometry. When we asked a professional lighting designer about using these simulations she indicated that although they are useful, "lighting designers really paint with light. What we'd really like to do is paint on the surfaces of the architectural model; then we work backwards, figure out where to put lamps to get the effect we want." [ref: personal communication, Erwine, 2002]

Light Pen is a working prototype that lets designers sketch with light. It uses sketching in 3D as input to a knowledge-based lighting design decision-support system. The designer specifies where the illumination should beis desired by sketching directly on surfaces in a 3D model, and Light Pen selects and , places and visually indicates the light sources and then visually indicates their effects. However, it doesn’t attempt to accurately represent the space lit. When that is desired, the designer can produce renderings by transferring the model (along with the fixtures and their locations that Light Pen has helped select) into an accurate light rendering program such as Radiance or Lights

cape.

The Light Pen system also illustrates a more general idea: using sketching and diagramming to interact with knowledge-based design systems. Earlier work explored this idea in two dimensions: posing queries to visual databases [ref] and setting scenarios for simulations [Gross and Do 2000ref]; here we extrapolate this work to sketching on a three dimensional model. Light Pen therefore comprises two components: SPenSpace Pen, a system we built earlier for annotating computer graphics models on the Web [xxx Jung 2001] and Lux, a simple knowledge-based expert system for lighting design decisions that we built built just to illustrate this idea. Figure 1 shows the Light Pen structure, with a main java Java applet including the SPen implementation and Lux. The light fixture database is an external file the applet loaded by the applets as needed. Other The system also interacts with other external files exist likesuch as the VRML description of the model or and the library of symbols that SPen can recognize.

Figure 1. Light Pen's internal structur

e

In the following sections we first describe a motivating scenario in lighting design and , then review related work in this area, outline the SPen annotation system and the Lux knowledge based design system, and then describe how the two parts work together as a whole system. We then outline the system and its different components: Space Pen to view and draw in 3D, the Geometry Analyzeryser module and the Lux knowledge-based system. We conclude with a discussion and identify directions for future work.

2 SCENARIO

Consider an architect designing a house. The architect designs and models the building using a conventional 3D modeler like Form•Z or ArchiCad and saves it in VRML format. The architect then imports the model into Light Pen to select and position lighting fixtures. After walking around the model the architect, deciding that the kitchen counter, as a working surface should be illuminated independently than from the rest of the kitchen, picks up the light pen and paints sketches on the counter top of the counter (figure 2).

Figure 21. The designer sketches light on top of the kitchen counter (right). The left frame indicates the designer’s point of view..

Consider an architect designing a house. The architect designs and models the building using a conventional 3D modeler like Form•Z or ArchiCad and saves it in VRML format. The architect then imports the model into Light Pen to select and position lighting fixtures. After walking around the model, the architect, deciding that the kitchen counter, as a working surface, should be illuminated independently from the rest of the kitchen, picks up the light pen and sketches on the counter top (figure 1).

The program registers the nearby surfaces and the size and shape of the light drawing sketch mark and could proposes 2 two solutions for illuminating the area. The first one is a series of 3 three spotlights directed toward the work surface and , aligned along the counter (Figure 32, left). The second solution is a pendant light close to the work surface illuminating mainly the center of the counter (Figure 2, right). ** later on I'd like to know how the system handles alternative design solutions and allows the architect to view them and select one. ** The architect chooses the first solution and the Light Pen system adds the lighting fixture to the model and diagrammatically indicates its illumination pattern.

Figure 32. Light Pen proposes a series of spotlights (left) or a pendant light (right) above the working surface..

The fixture appears on the floor plan as well. EventuallyIn a future version, theThe architect could would then request energy consumption and pricing information about the fixtures that Light Pen has selected, from a manufacturer-supplied database. Other solutions can be found for illuminating the entire kitchen area or a painting on the wall. Each time the system selectsfinds appropriate lighting fixtures according to the geometry of the surfaces that the light is painted on.

Our first prototype of Light Pen can already choose and display a few lighting fixtures. Multiple options and energy or pricing information are still to be implemented.

3 Related work

Since the early 1970s computer graphics research has constantly continually improved techniques to accurately render lighting effects in computer generated images. Today products such as Radiance or and Lightscape achieve excellent simulation of either both daylighting orboth daylighting and artificial lighting effects. However, rendering such images is still processor-intensive and time-consuming. Renderings Accurate renderings still cannot be generated in real time; therefore they can only be produced for still images or movies. Some research teams continue to work to Efforts are underway to optimize the light simulation process (e.g., by parallelization) in order to perform accurate renderings for real-time interactive 3D worlds [Robertson, 1999].

Light Pen operates in a real-time 3D environment. The current version of Light Pen provides diagrammatic illumination rendering to illustrate the characteristics of a lighting solution rather than high fidelity rendering effects. Eventually, however, as in Light-Sketch [Glaser 2003], Light Pen could connect with high fidelity simulators to render a more accurate representation of the space with proposed alternative lighting solutions. LightSketch [Glaser 2003] is a 2D interface that enables a user to sketch light symbols (electric lighting and windows) and render the effects using a third party renderer such as RADIANCE.

Light Pen operates in a real-time 3D environment, but its lighting representation does not aim for high fidelity; rather Light Pen's diagrammatic illumination rendering merely aims to help a designer quickly grasp the lighting characteristics of a solution. (After all, in Light Pen the designer specifies the desired illumination in the first place). Eventually, however, Light Pen could work together with high-fidelity simulators to render a more accurate representation of the space with the alternative lighting solutions it has proposed.

In Light Pen's 'reverse lighting calculation' the designer specifies the position of the illumination, rather than the position of the light sources. Others have also followed this approach. Schoeneman and Dorsey's The "Painting with Light" system [Schoeneman et al., 1993] calculates the intensity and color of a light source (at a given position) needed to produce a desired illuminating effect that the user paints on the surfaces of a 3D model. It then produces a precise representation of the space with the calculated light features. However, that system is designed only to calculate intensities and colors for theatrical lighting with fixed position light sources. Light Pen, on the other handin constrastcontrast, uses a simple knowledge based system to select and locate appropriate light fixtures in an architectural spacea building.

Poulin et al. [Poulin, 1997] sketch shadows and highlights on a 3D model to find the position of the most appropriate light source. However, their system has only one kind of light (a basic point light) and the position of the light source is mainly determined by the size and location of the shadow. There isIt has no knowledge based system or fixture database behind it. In the same vein the system of Ayatsuka et al. [Ayatsuka, 1996] positions objects (spheres) in 3-D given the location of shadows produced by three fixed light sources. Light Pen uses sketched highlights, the surrounding architectural geometry and a preset light database to propose the types and positions and lighting typesof architectural lighting .

Light Pen is a graphical interface to a knowledge-based system on for lighting design. Knowledge based expert assistants for the engineering aspects of architectural design were the topic of considerable research in the 1980s. F, for example: the HI-RISE supported the selection and dimensioning of structural systems [Maher 1985], and SOLAREXPERT supported solar design tasks [Rosenman 1987] and energy and. thermal behavior of buildings [Brown1995]. Most suchThese systems, however, were not integrated into a 3D sketching environment. In lighting design, products such as EcoLumens [5www.tatainfotech.co.in] or and LightCalc [6www.lightcalc.com] propose alternative lighting of a whole building. These programs' knowledge systems are based on tasks or energy efficient considerations. ** what does that mean? ** You specify the size of your (rectangular) room, tell the system what kind of room it is (bedroom, hall, office, etc.) and the system displays a solution. You can then adjust by specifying energy consumption or cost, and the system will propose a new lighting fixture. Solutions for placing the lights are limited to a few simple options.

1 They work on one (rectangular) room at a time with a simplistic 2D representation of the space and illumination solutions.

4 SPenSystem overview

Light Pen is a Java applet bringing together Space Pen in Java3D and Lux, which is implemented in Java. The virtual model is in VRML format and imported in Java3D through the VRMLLoader class; a Java library for VRML objects developed by the x3d task workgroup. Once converted to Java3D, the system must prepare the model to be drawn on by activating the picking capabilities of all its surfaces.

Figure 3 shows the internal structure of the Light Pen system, with its three main components: Space Pen, to import the 3D VRML model and draw on it, the Geometry Analyzeryser module to query about the model’s surfaces, and Lux, the knowledge based system that will choose and position light fixtures.

Figure 3. The Light Pen components, including the Geometry Analyzeryser module.

Let us begin by explaining how Light Pen understands a “light object” (the sketch mark that the designer makes to indicate where light is desired) and the surfaces on which that light object is drawn. When a user marks on the model with the Light Pen, the system saves the geometry of the mark (coordinates, normals, size, and ink) and registers all the surfaces on which the mark has been made. Light Pen then analyses the registered surfaces in relation to each other and to surrounding surfaces. One might think that only seldom does a drawing touch more than one surface. However, if the user wants to indicate that an entire room is to be illuminated, it is likely that several surfaces of the room will be painted with the light pen. Also, when the model is transferred into VRML it is usually triangulated, so what appears as a single surface in the 3D environment is likely to be composed of several small triangulated surfaces. However, in most cases the registered surfaces for each light object can be reduced to a small set of coplanar ones.

Lux is a knowledge-based system for choosing and placing a lighting fixture. It demonstrates in principle how aspects of lighting design expertise can be represented in a set of condition-action rules. However, it is a simplification of a real lighting designer's expertise. The Lux expert system takes into account both the geometry of the light object and different aspects of the registered surfaces to determine which light fixture to deploy and how to deploy it.

4.1 Space Pen

Light Pen is built on top on a previous system we built called SPenSpace Pen. SPenSpace Pen is a collaborative tool that supports sketching on 3D web models. Architects and their clients communicate by sketching annotations on a 3D virtual model posted on the SPenSpace Pen web site. Clients and other team members can view their project in 3D before it i's built and leave comments as post-it notes or drawings directly within the 3D environment. Later, designers can review those comments and modify the project accordingly.

SPenSpace Pen is above all an annotation system. The user can draw objects and symbols on any surface of the model. Sketched annotations can beare saved in VRML format for future reviewers to see. Post-it comments are indicated in the model as yellow boxes and a visitor can participate in a threaded discussion by clicking on a comment and filling typing in the reply text field.

** this paragraph and the following sentence about 'drawing on existing surfaces - is interesting and explains what Space Pen does, but it doesn't seem to have anything to do with the Light Pen system. Is there any way to make this paragraph more meaningful in the context of the current paper? ** SPen can recognize, (and therefore "beautify") symbols such as rectangles, circles, triangles or arrows. This is used mainly to represent more accurately the visitor’s intention, as it is still difficult to draw with a pen tablet or a mouse. However, in a future implementation, each recognized symbol may be associated with a specific command within SPen. For example, drawing an arrow next to an object could move that object.

Sketches in SPen are made on existing surfaces, or in the space by first generating a temporary drawing surface. Space Pen serves also as a navigation browser in a 3D environment. It automatically computes and displays a floor plan of the space and indicates the user’s position and orientation. Light Pen uses Space Pen’s environment and capabilities to navigate in the 3D virtual model.

Navigating in a 3D space can be challenging for novice users, and even for more experienced designers. None of the existing VRML browsers were adequate, so we built a simple 3D browser into SPen using the Java3D classes that support VRML data format. To facilitate navigation through a 3D virtual world, SPen indicates the visitor's position in the floor plan, displayed on the left frame of the window. The floor plan is calculated directly from the 3D model and as the visitor moves up or down SPen regenerates the plan at the current visitor's eye level.

The visitor's location is saved as a viewpoint with each new annotation or comment. This provides another way to quickly navigate through the model, by clicking on each viewpoint marker on the floor plan window.

In The Light Pen the system adds a viewpoint with every lighting fixture it proposes. The designer can use these viewpoints to quickly find and examine the added lighting fixtures.

Lux

Lux is a simple knowledge-based system for lighting design decisions. We built Lux only to illustrate how one might link the SPen to an expert system. Lux takes into account the geometry of a 3D model to make decisions about artificial lighting. The first implementation of Lux is linked to a database of light fixtures to choose from. The light fixture database is a separate text file containing a list of available lights with their characteristics such as type, intensity, color and; these could eventually be extended to include cost or and energy consumption.

Each time the designer makes a light mark on a surface, Light Pen analyses its geometry and its relation to surrounding surfaces. Each surface is given labeled with a type, such as work surface, ceiling, wall or floor. Lighting designers seek to provide a lighting level appropriate to the activity that will take place in the space. The IESNA Lighting Handbook [ref] identifies three different type of lighting: Ambient, Task and Accent. Ambient lights concern a whole space, task oriented lighting is designed for highlighting a certain work area, whereas accent lighting draws attention to a certain object like such as a painting or sculpture. According to the surface type and to the light sketch, Lux determines what type of lighting is needed (ambient, task or accent) and then proposes an appropriate lighting fixture to illuminate the chosen surface.

Lux also proposes locations for the lighting fixtures considering the position of the illumination sketch in the model and the availability of “ceiling surfaces” (or surfaces considered as ceilings) nearby.

SYSTEM OVERVIEW

Light Pen is a Java applet bringing together SPen in Java3D and Lux, which is implemented in Java. The virtual model is in VRML format and imported in Java3D through the VRMLLoader class; a Jjava library for VRML objects developed by the x3d task workgroup. Once converted to Java3D, the model must be prepared for drawing on it by activating the picking capabilities on of all the its surfaces.

The configuration of the space and the user's intentions on where illumination should be are two important inputs to the lighting design’s decision-making process. Expert lighting designers also base their decisions on additional considerations, such as the building purpose (a hospital room is not lit like a private house bedroom or a classroom) or the surface materials used. However, the current version of Lux considers mainly the surrounding architectural geometry of the light sketch, but future work will explore that area.

Lux is a simple knowledge-based system for choosing and placing a lighting fixture. It demonstrates in principle how aspects of lighting design expertise can be represented in a set of condition-action rules. However it is a simplification of a real lighting designer's expertise.

Let us begin by explaining how Light Pen understands a “light object” (the sketch mark that the designer made to indicate where light is desired) and the surfaces on which that light object is drawn. When a user marks on the model with the Light Pen, the system saves the geometry of the mark (coordinates, normals, size, and ink) and registers all the surfaces on which the mark has been made. Light Pen then analyses the registered surfaces in relation to each other and to surrounding surfaces. One might think that only seldom does a drawing touch more than one surface. However, if the user wants to illuminate an entire room, it is likely that several surfaces of the room will be painted with the light pen. Also, when the model is transferred into VRML it is usually triangulated, so what appears as a single surface in the 3D environment is likely to be composed of several small triangulated surfaces. However, in most cases the number of registered surfaces for each light object can be reduced to a small set of coplanar ones.

Lux is a knowledge-based system for choosing and placing a lighting fixture. It demonstrates in principle how aspects of lighting design expertise can be represented in a set of condition-action rules. However, it is a simplification of a real lighting designer's expertise. The Lux expert system takes into account both the geometry of the light object and different aspects of the registered surfaces to determine which light fixture to deploy and how to deploy it.

The configuration of the space and the user's intentions of where to illuminate it are two essential inputs to the lighting design’s decision-making process. Expert lighting designers also base their decisions on additional considerations, such as the building purpose (a hospital room is not lit like a private house bedroom or a classroom) or the surface materials used. However, the current version of Lux considers mainly the architectural geometry surrounding the area to be illuminated.

The Lux expert system takes into account both the geometry of the light object and different aspects of the registered surfaces to determine which light fixture to deploy and how to deploy it.

4.2 The “Geometry Analyzeryszer” module

Light Pen analyzes some characteristics of the light sketch the designer drawsn (its shape, size, location, etc.) and as well as the configuration of the surrounding surfaces. The goal is to identify every surfaces composing the model, not only by its their coordinates and appearance but also by its their functions and common names. For example, Light Pen would recognize a vertical surface with a transparent material as a window. ** Can it do this now? How hard to do this? It would be nice...** A horizontal surface placed at a height higher than 0.7 meters (and lower than 1.2 meters) from the floor is likely to be a desk or some kind of work surface. By knowing identifying the kind type of object the user is drawing on, the Light Pen system can make better suggestions about what light fixture to choose.

Among the surface characteristics evaluated by the Geometry Analyzeryszer Module are height, size, orientation and coplanarity of the surface, existence of a ceiling nearby and distance to the ceiling Analyzer module are co planarity, orientation, existence of a ceiling nearby, distance from the surface to the ceiling, and height and size of the surface. The GeometryThe Geometry Analyszer module returns boolean values that Lux subsequently uses to infer the appropriate lighting solution. Each surface characteristic is calculated in its own method and every time a the designer draws a new light mark is drawn.

4.3 Knowledge Based System: Lux

Lux is a simple knowledge-based system for lighting design decisions. We built Lux only to illustrate how one might link the Space Pen to an expert decision-support system. Lux takes into account the geometry of a 3D model to make decisions about artificial lighting. The first implementation of Lux includes a database of light fixtures to choose from within the applet. In the next version, the light fixture database would be a separate text file containing a list of available lights with their characteristics such as type, intensity or color.

Each time the designer makes a light mark on a surface, Light Pen analyses its geometry and its relation to surrounding surfaces. It labels each surface with a type, such as work surface, ceiling, wall or floor. Lighting designers seek to provide a lighting level appropriate to the activity that will take place in the space. The IESNA Lighting Handbook [IESNA 2000] identifies three different types of lighting: Ambient, Task and Accent. Ambient lights concern a whole space, task oriented lighting is used to highlight a certain work area, whereas accent lighting draws attention to a certain object such as a painting or sculpture. According to the surface type and the light sketch, Lux determines what type of lighting is needed (ambient, task or accent) and then proposes an appropriate lighting fixture to illuminate the chosen surface.

The configuration of the space and the user’s intentions of where to illuminate it are two essential inputs to the lighting design’s decision-making process. Lux Lux uses a set of decision rules to first identify determine which lighting fixture to use and then its position. The first rule concerns the co coplanarity of the registered surfaces. If the surfaces are not coplanar, the system infers that it is a global larger space that needs to be lit. It must then find an ambient lighting system capable of lighting an entire space. It can probably eliminate direct spotlights from the list of possible fixtures, as spotlights being are more often used as for task or accent lighting.

To better understand the rest of the knowledge system let’s consider the scenario above where the user wants to illuminate a kitchen counter (figure 2). Figure 4 (below) shows the decision tree that drives the current Lux implementation. The arrows along the tree show the decision process for our scenario of lighting the kitchen counter.

As before, the system first determines that the surfaces on which the user has drawn (the counter) are coplanar. Then Light Pen checks the orientation of the surface. (A vertical surface won’t require the same illumination as a horizontal one.) The counter is horizontal. Light Pen then checks whether there is a ceiling (or any horizontal surface) above the counter. If there is one (and in our scenario there is), it checks if the surface is in a small enclosed space by calculating how far the ceiling is from the surface. The kitchen counter doesn’t qualify so Light Pen then checks the height of the registered surfaces relative to the surface on which the user is standing (which it assumes to be the floor). In our case the illumination sketch was made on a kitchen counter 1 meter high from the floor, which registers it as a work surface. In Light Pen’s knowledge base, illuminating a work surface calls for a task light, so a task lighting fixture will be needed. The last check concerns once again the distance to the ceiling. Light Pen makes sure that the distance between the counter and the ceiling is not longer than the maximum distance where the light fixture would be effective. Because that distance is within the range of what, at least, one lamp could illuminate, Light Pen could suggest mounting either a spotlight or a fluorescent light under the cabinet above the counter surface.

Figure 4. The decision tree used in the first implementation of Lux.As before, the system will first determine that the surfaces on which the user has drawn (the counter) is are coplanar. Then Light Pen checks the orientation of the surface. (A vertical surface won’t require the same illumination as a horizontal one.) The counter is horizontal. Light Pen then checks if theirs there is a ceiling (or some kind ofany vertical horizontal surface) above the counter. If there is one (and in our scenario there is), it calculates the distance from the counter to the ceiling to determine if it is a small- enclosed space. The kitchen counter doesn’t qualify for that so Light Pen then checks the height of the registered surfaces relative to the surface on which the user is standing (and therefore, most probably, thewhich it assumes to be the floor). In our case the illumination sketch was made on a kitchen counter 1 meter high from the floor, which registers it as a work surface. In Light Pen, illuminating a work surface mean calls for a task light, so a task lighting fixture will be needed. The last verification check concerns once again the distance to the ceiling. Light Pen makes sure that the distance between the counter and the ceiling is not longer than the maximum distance where the light fixture would be effective. Because that distance is within the range of what, at least, one light lamp could reachilluminate, Light Pen will suggest hanging mounting either a spotlight or a fluorescent light under the cabinet over above the counter surface. Both alternatives can later be displayed graphically in the 3D model.

Figure 4 (below) shows the decision tree that drives the current Lux implementation. This structure can be easily changed in the code to reflect a more sophisticated lighting design decision process.

Figure 4. The decision tree used in the first implementation of Lux.

Looking at tThe Lux decision tree shows that , each type of light fixture could may be described chosen by a basic set of rules on the surrounding geometry. Table 3 1 shows the set of rules conditions that drives Lux to choose a pendant light over another type of light fixture. There is Cucurrently two different sets of rules that will that both lead to a pendant light.: One is for a large illuminated areaion on the floor and in casewhen the above ceiling above is higher than the maximum reach throw of other light fixtures. The other set of rules include respond to the need for a big large illumination sketched area over non-coplanar surfaces and with an existing ceiling above.

Once Light Pen has decided which fixture to use, it must decide where to place it. This depends, of course, on the fixture itself, but also on some aspects of the geometry. For most cases (spotlights, pendant lights fluorescent lights), Light Pen finds the centre of the illumination sketch, projects it up toward the ceiling, and places the light source there. ** the rules as they are written are difficult for a reader to understand. Either we need to explain (walk the reader through) the rules in the text, or add comments to the rules. Also it might be a helpful idea to walk the reader through an example in the decision tree, and maybe even highlight (or darken the lines of) a path through the tree that we explain in the text. **

Table 31. The decision rules for choosing a pendant light in Lux

Table 4 sum upshows the rules for Light Pen to choose a spotlight kind of fixture. Spotlights are chosen over other type of lighting fixtures mostly when the illumination area is small or when the space where the light should be to be illuminated is enclosed.

Table 4. The decision rules for choosing a spotlight in Lux

Light Pen combines a sketching interface (SPen), a knowledge-based system (Lux) and a database of light fixtures. Although the implementation is at a rathern early preliminary stage, and the rules of the knowledge system are reflecting only a small part of the decision process of a lighting designer, Light Pen demonstrates the possibility of linking a how an expert system can be linked to a graphical user interface, all that onin a java applet, running over the Web.

THE NEXT PARAGRAPH IS IN FUTURE TENSE AND THEREFORE NOT IMPLEMENTED YET (BUT IS IT CLEAR ENOUGH ? DOES IT NEED TO BE CLEARER ?):

Light Pen combines a sketching interface (Space Pen), a knowledge-based system (Lux) and a database of light fixtures. Although the implementation is at a preliminary stage and the rules of the knowledge system reflect only a small part of the decision process of a lighting designer, Light Pen demonstrates how an expert system can be linked to a graphical user interface, all in a java applet, running over the Web.

** We've explained how the system selects a lighting fixture, but not how it decides where to place the fixture. Can this be briefly explained here?

5 DISCUSSION and FUTURE work

Future developments of light Light pen Pen could may go develop in several directions. We could further develop further the knowledge-based system to reflect more accurately reflect the decision making of a lighting designer. OIn that case, other elements must then be taken into account: The the material of the surrounding surfaces (reflective surfaces, absorbing surfaces, etc.), the location of the openings (windows, glass doors) or/and the building type. Buildings can be categorized by their purpose, size, location, each of which would have an influences on the lighting decisiondesign: The illumination of a hospital in Seattle does not follow the same ruleshas different requirements than from a classroom in Casablanca or a private home in Moscow.

We are currently also considering how to take into account the effects of daylighting on a 3D model. We would like our system to identify the skylights, the windows and the sun path in the model in order to make more informed decisions onabout the light fixtures and their locations. Dan Glaser’s project [Glaser 2000] analyzes daylight inputs in a building and displays results in a two-dimensional representation of the space.

Finally we could improve the Light Pen capabilities by being able to not only specify the location of the illumination but also the intensity or/and color of the light. In that matter, we would not only sketch with light, but paint with light as Schoeneman and Dorsey [Schoeneman et al. 1993] proposed. Linked to our knowledge base the system would be able to choose different fixtures and place them automatically.We would like to extend our Geometry Analyzer module to recognize and characterize each every surface of a randomny given building model of a building. At the current stage of our implementationCurrently only a few surface type can be identifyidentified: Floor, ceiling, walls (as vertical surfaces) and work surfaces. Windows, furniture, stairs and other random elements are not yet differentiated. Having the model described not only geometrically, but also semantically would certainly simplify not only the light paint algorithm but also other functions in SPen ** such as?** . More research needs to be done in how to differentiate every element of a building using only geometrical and material or texture information (the only information available in VRML format). ** it is an interesting and 'nice' problem to parse the 3d model for architectural semantics, but it is also a large problem and it is quite beyond the scope of LP. And, it may be that the next generation of 3d modeling software for architecture will provide semantic labels to the objects and surfaces. In that case, building to get around VRML's limits is silly. So - this 'future work' seems interesting but I wonder how seriously we want to discuss it here as something we 'plan' to do.

** painting (rather than sketching) with light : (a) setting the intensity of illumination (and illumination patterns), (b) setting the color of illumination

** taking into account daylighting as well as artificial lighting (windows, skylights), adding this to the knowledge based system

References

Ayatsuka, Y., S. Matsuoka and J. Rekimoto. 1996. Penumbrae for 3D Interactions, Proceedings of the ACM Conference on User Interface Software and Tools (UIST) (TechNote),,: ACM Press, pp.165-166.

BlinnBlinn, J. F. and, M. E. Newell, M. E. .,1976. Texture and reflection in computer generated images:, Communications of the ACM, October 1976, volume 19, Issue 10, pp. 542-547.

Brown, G.Z., Kline, J. and Sekigitchi, T., Infrared Professor - Design Phase, Sixth International Conference on Computer-Aided Architectural Design Futures [ISBN 9971-62-423-0] Singapore, 24-26 September 1995, pp. 103-112

Erwine, B.. April 2002. Personal Communication.

Glaser, D. and S. Ubbelohde. 2001. Visualization for Energy Efficient Building Design, Extended abstracts of the IEEE Symposium on Information Visualization

Glaser, D., J. Voung, L. Xiao, B. Tai, S. Ubbelohde, J. Canny, E. Y.-L. Do. 2003, LightSketch: A sketch-modelling program for lighting analysis. CAAD Futures 2003, Tainan, Taiwan, Kluwler.

Greenberg, Donald P., 1974. Computer Graphics in Architecture:, Scientific American, vol. 230, no. 5, May 1974, pp. 98-106.

Gross M. D. and E. Y.-L. Do. 2000. Drawing on the Back of an Envelope: a framework for interacting with application programs by freehand drawing: Computers and Graphics, Vol. 24 (6) pp. 835-849.

Jung T., M. D. Gross and E. Y.-L. Do. 2001. Space Pen: Annotation and Sketching on 3D Models, CaadAAD Futures pProceedings, pp. 257-270.

Maher M. L., i-Rise and Beyond : Directions for Expert Systems in Design, Computer Aided Design. November, 1985. vol. 17: pp. 420-427Maher, M. L. 1985. HI-RISE and beyond: directions for expert systems in design:, CAD Journal, 17(9) (1985), pp. 420-427.

Newell,, M. E. and, J. F. Blinn J.F. ,1977. The progression of realism in computer generated images, Proceedings of the 1977 aAnnual ACM/CSC-ER conference, 1977, pp. 444-448.

Poulin, P.,; K. Ratib and M.b, K.; Jacquess, M., 1997.(1997), Sketching shadows and highlights to position lights, IEEE Proceedings of IEEE Computer Graphics International Cconference, 1997. Proceedings , pp. 56 -63.

Robertson D., Campbell K., Lau S., Ligocki T., Parallelization of Radiance For Real Time Interactive Lighting Visualization Walkthroughs, Proceedings of the 1999 conference on supercomputing 1999, pp 60-68

Rosenman, M. A., J. S. Gero, J. S., R.D. Coyne R. D., et al. , 1987. SOLAREXPERT : A Prototype Expert System for Passive Solar Energy Design in Housing., Canberra: Australia- NZ New Zealand:, Solar Energy Society, 1987. vol.II: pp. 361-370.0

Rosenman, M. A., Application of Expert Systems to Building Design Analysis and Evaluation, CADline proceedings, Building and Environment. 1990. vol.25: pp. 221-233

Schoeneman, C., J. Dorsey, B. J., Smits, J. B., Arvo J. and D., Greenberg D. 1993., (1993), Painting with Light, Proceedings of the 20th annual ACM Cconference on Computer graphics and interactive Interactive techniquesTechniques, (SIGGRAPH )93, 1993 pp. 143-146.

xxx., yyy., zzz, Spen: Annotation and Sketching on 3D Models, aaa…

Lighting Fundamentals, Lighting Upgrade Manual, US EPA Office of Air and Radiation 6202J, 199

IESNA. 2000. Lighting Handbook - 9thEdition: Illuminating Engineering Society of North America., 2000

Ecolumen: www.tatainfotech.co.in

LightCalc: www.lightcalc.com