Virtual Alpine Landscapes and Autonomous Agents Duncan CAVENS, Eckart LANGE and Willy SCHMID 1 Introduction For most landscape architects, the concept of computer simulation stops at the creation of a 3D simulation of a design proposal. While visual simulation is important for exploring design options and communicating with a wider audience (SHEPPARD 1989), a wider use of other forms of modeling and simulation to explore the impacts of design interventions could give designers and planners another tool to help accomplish their tasks. One key issue that has hampered the adoption of simulation in small scale planning and design projects is the sheer complexity of the associated issues, particularly with respect to the social impact of proposed interventions. This paper introduces the project “Planning with Virtual Alpine Landscapes and Autonomous Agents”, which is funded by the Swiss National Foundation program “Habitats and Landscapes of the Alps.” The project is exploring the feasibility of using autonomous agent modelling to evaluate proposed changes to an alpine landscape. The project seeks to use simulated people (agents) who “see” the landscape as surrogates for real people reacting to the proposed future landscapes. This paper describes the overall project approach, and explains how visualisation will be used in the context of the project. 2 Visual Perception Research Although visual concerns are often considered too vague to be included in a simulation (which implies being able to quantify them), there is a long history of using photographs and visual simulations to quantify the visual quality of the landscape. The standard technique is to have individuals assess photographs of a particular location, and use their ratings to determine the visual quality of a view or landscape (e.g. SHAFER AND BRUSH 1977). Many studies (e.g. DUNN 1976, STAMPS 1993, MEITNER AND DANIEL 1997) have demonstrated that evaluations made on the basis of photographs correlate closely with people’s assessments of the places they represent. Studies have investigated the impact of various features (such as water, percentage of vegetation, types of tree shapes) on people’s aesthetic response to a particular scene(e.g. SUMMIT AND SOMMER 1999) A refinement of using photographs to evaluate landscape quality is to use computer simulations (LANGE 2001). Using computer graphics, subjects are shown a series of computer generated images in which minor elements are altered. The impacts of these changes are measured with respect to the overall degree of realism of the scene. This allows the researcher to more accurately judge the impact of specific elements and/or
56
Embed
Virtual Alpine Landscapes and Autonomous Agents · 2010. 1. 10. · Virtual Alpine Landscapes and Autonomous Agents Duncan CAVENS, Eckart LANGE and Willy SCHMID 1 Introduction For
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Virtual Alpine Landscapes and Autonomous Agents
Duncan CAVENS, Eckart LANGE and Willy SCHMID
1 Introduction
For most landscape architects, the concept of computer simulation stops at the creation of a 3D simulation of a design proposal. While visual simulation is important for exploring design options and communicating with a wider audience (SHEPPARD 1989), a wider use of other forms of modeling and simulation to explore the impacts of design interventions could give designers and planners another tool to help accomplish their tasks. One key issue that has hampered the adoption of simulation in small scale planning and design projects is the sheer complexity of the associated issues, particularly with respect to the social impact of proposed interventions.
This paper introduces the project “Planning with Virtual Alpine Landscapes and Autonomous Agents”, which is funded by the Swiss National Foundation program “Habitats and Landscapes of the Alps.” The project is exploring the feasibility of using autonomous agent modelling to evaluate proposed changes to an alpine landscape. The project seeks to use simulated people (agents) who “see” the landscape as surrogates for real people reacting to the proposed future landscapes. This paper describes the overall project approach, and explains how visualisation will be used in the context of the project.
2 Visual Perception Research
Although visual concerns are often considered too vague to be included in a simulation (which implies being able to quantify them), there is a long history of using photographs and visual simulations to quantify the visual quality of the landscape.
The standard technique is to have individuals assess photographs of a particular location, and use their ratings to determine the visual quality of a view or landscape (e.g. SHAFER
AND BRUSH 1977). Many studies (e.g. DUNN 1976, STAMPS 1993, MEITNER AND DANIEL
1997) have demonstrated that evaluations made on the basis of photographs correlate closely with people’s assessments of the places they represent. Studies have investigated the impact of various features (such as water, percentage of vegetation, types of tree shapes) on people’s aesthetic response to a particular scene(e.g. SUMMIT AND SOMMER
1999)
A refinement of using photographs to evaluate landscape quality is to use computer simulations (LANGE 2001). Using computer graphics, subjects are shown a series of computer generated images in which minor elements are altered. The impacts of these changes are measured with respect to the overall degree of realism of the scene. This allows the researcher to more accurately judge the impact of specific elements and/or
D. Cavens, E. Lange and W. Schmid 2
spatial arrangements; for example if a tree is in the foreground, it could potentially increase the visual quality of a scene.
Most of the above research into landscape perception has largely been descriptive: it investigates how people react to a particular landscape or elements thereof, but often falls short of interpreting these reactions and exploring how they can be used to improve the quality of landscape planning and design. While this knowledge is useful for informing design decisions, conducting these kinds of tests with real subjects is time-consuming and is generally not considered outside of research environments. It is not considered practical as a technique for evaluating different design proposals for typical projects.
Our project explores an alternative approach: rather than using real human subjects to evaluate potential design scenarios, computer simulations of individuals are used to evaluate the proposed changes to the landscape. By using data gathered using standard perception testing techniques, it is hoped that these simulated individuals can be calibrated to represent real individuals in the real landscape.
3 Test Site: Schönried, Switzerland
The specific test site is a valley in the Gstaad-Saanenland region of south-western Switzerland. The communities of Schönried and Saanenmöser are at the two ends of the site; their economies are highly tourism dependent. While the primary tourism draw to the area used to be winter skiing, long term climate change is forcing the community to focus its efforts on building up a more diversified tourism economy. This includes capitalizing on its already strong reputation for summer hiking. The landscape is a mixture of pasture and coniferous forests, dominated by Norway Spruce (Picea abies). The test site is characterised by significant topography and is considered ideal for walking and hiking. (see figure 1). The trails are very accessible to a wide range of hiking abilities due to the summer operation of one chair-lift and two gondolas. In the high season, the area is busy with hikers and walkers who easily fill the two main parking lots in Schönried.
A recent study in the area (MÜLLER AND LANDES 2001) identified that the biggest attraction for summer tourists are the area’s scenic qualities. Hiking and walking is the primary recreational activity in the summer months. The focus on views was confirmed by our own study in 2002 (CAVENS AND LANGE 2003), where views and landscape variety were identified as the most important factors that influenced hikers in their choice of hiking routes.
In addition to the community’s desire to diversify its recreational economy, there are landscape policy issues that have the potential to change the desirability of the area for summer tourism. These issues include changes to the pattern of the landscape due to changing agricultural policy, shifts in forestry practices, closing of the gondolas and/or chairlifts, and increased holiday home construction. Any of these changes would have complex repercussions for the tourism industry: future scenarios to test the agent model will be selected from them.
Virtual Alpine Landscapes and Autonomous Agents 3
Figure 1: Typical view of study site (from Horneggli mountain towards Schönried )
4 Modeling Approach
Our project couples realistic 3D computer visualisation with autonomous agent modeling to test alternative landscape scenarios. While the 3D visualisation is useful as a tool to explore the model and confirm its correctness, its primary purpose is to confirm, with real subjects, that the simulation’s assumptions about visual and aesthetic concerns are accurate.
Obviously, there are certain classes of landscape problems that are better suited to this approach than others. The use of simulation in planning has largely been limited to investigating questions that affect large areas, usually at the regional or larger scale. These planning related simulations have either dealt with wildlife or ecology issues (HEHL-LANGE
2001) or large scale urban systems problems (such as traffic congestion (NAGEL 1998) or the optimal allocation of land uses (WEGENER AND FÜRST 1999)). At these scales visual concerns have little impact on the phenomena being modelled, and are safely disregarded. However, other than overall regional strategies, most planning decisions are made at a much finer scale than the available simulations. These decisions, such as those to allow an increased density of housing around a village centre, have a direct impact on local residents that is not easily captured by city-wide or regional models. At these small scales (such as the sub-watershed or village), visual elements and the overall visual quality of the proposed planning intervention are extremely important. This is particularly true with areas
D. Cavens, E. Lange and W. Schmid 4
dependent on tourism, which are often promoted based on their scenic qualities. The impact of changes is often cumulative, as the overall impact of a series of design projects is larger than the individual impact.
For our particular site, the central question is to determine what impact changes to the landscape have on the ‘enjoyment’ of summer hikers, and will those changes have enough of an impact for them to either change their typical route, or, decide not to return to the same place next year? This is a very difficult question to answer, because changes to the landscape (unless major) have the possibility of affecting individuals in very different ways.
Autonomous agent modelling allows one to represent each resident or visitor in the simulation, rather than be forced to aggregate their preferences into a general model. Each ‘agent’ is given a set of preferences and goals and is set loose to explore the environment. As they move through the environment, they learn from their experiences. Eventually, if the parameters are specified correctly, the goal is to have a simulation which reflects the current observed situation (i.e. agents move in the same places and at the same times as in the real world.). Once this has been achieved, changes to the environment can be introduced, and the agents’ reactions to the model studied.
Others have applied autonomous agents to small scale landscape planning concerns (GIMBLETT et al. 2002). The research so far has been confined to two dimensions. While these simulations have been applied to recreation areas, their interest has primarily been on how users interact with each other, rather than on how people interact with the landscape itself. While these ‘traffic’ issues are important in the busiest of recreation sites or in wilderness areas where encounters with other hikers are not typically desired, they are not the critical concern in the majority of recreation communities where retaining and attracting more tourists is often more important than managing those that already visit.
Simulating individuals allows one to create a highly nuanced model of visual responses, with each having a different tolerance for different kinds of visual stimuli. However, simulating the activities and visual preferences of thousands of individuals is a complex task, and requires both a large amount of source data and ample available computing resources. We are relying on the agent modelling framework developed by our project partners at ETH’s Institute for Scientific Computing (Prof. Dr. Kai Nagel). Their system, based on distributed computing concepts, provides us with the ability to simulate millions of agent-days quickly, allowing one to tweak model parameters continuously.
Real People
/ Real World
Real People
/ Virtual World
Virtual People
/ Virtual World
Calibration
Virtual Alpine Landscapes and Autonomous Agents 5
Figure 2: Diagram of the overall project structure.
In order to produce realistic results, one requires a good set of base data to drive the agent simulation. The agent simulation relies on a GIS-derived spatial database, representing detailed spatial information such as the location of fences, individual trees, types of ground cover and the condition of trails and roads. The agents are ‘aware’ of these objects as they move through the landscape, and they use them to decide where to go, as well as inputs for their evaluation functions which determine how satisfied each agent is with their chosen route.
A detailed spatial model is useless for agent modeling if one is unable to calibrate the agents’ reaction to it. In order to accurately calibrate and validate our model, we are collecting data from respondents in two tracks: using formal surveys and informal observations out in the site itself; and will be using a virtual environment to track subjects’ response to the proposed changes to the virtual world. As illustrated in figure 2, data collected from the ‘real people in the real world’ and ‘real people in the virtual world’ will be used to confirm the results from the autonomous agent simulation (‘virtual people in the virtual world’).
Data regarding ‘real people in the real world’ is being collected through interviews with hikers on site, using counts of hikers using the various public transit facilities (trains, gondolas, chair lifts), and based on structured observation of usage patterns in the landscape. This data will provide an understanding of the current situation. By interpreting this data with a detailed GIS analysis of the site itself, we expect to be able to develop a model that reflects the visual preferences of the site visitors.
It is not necessarily easy to extrapolate from this understanding of the current situation to proposed landscape scenarios: in order to ensure that our agent simulation accurately reflects hikers sensitivities to change, subjects will be asked to evaluate these changes in a virtual environment (see BISHOP ET AL. 2001 for a discussion of the strengths and weaknesses of this approach). Subjects will be invited to explore a computer generated representation of the study site, both in its current state and with the proposed landscape scenarios applied. Their path choices will be compared and evaluated, and used to further refine the autonomous agent model.
5 Use of Visualisation
In order to test real people in a virtual world, the virtual world needs to first be created. As Bishop points out, due to limitations in current rendering technology (and available resources for modeling), it is not yet feasible for a virtual world to act as a complete surrogate for reality. However, as other research has demonstrated (LANGE 2001, DANIEL
AND MEITNER 2001), realistic renderings using current technologies are sufficient for subjects to make evaluations based on visual criteria. While these studies have been done on static images rendered from a 3D model, it is our belief that they also apply to virtual environments where users can move freely, as long as a minimum standard of interactivity and degree of realism is achieved.
D. Cavens, E. Lange and W. Schmid 6
The 3D rendering system serves two purposes in the project: to provide an environment for the testing of real people and to allow interaction with the agent based-simulation for testing and evaluation purposes.
Both purposes have similar requirements for the rendering system: being able to move dynamically though the landscape. The system has to be able to track movements (in the case of testing human subjects), and display the activities of computer generated agents. The following criteria were established for selecting a visualisation system and renderer:
• provide real-time rendering capability • provide highly realistic rendering • scale to accommodate future hardware developments and constantly-improving
rendering techniques • multiple platforms (from Windows laptops to SGI Onyx class machines) • to ensure consistency between models, be able to use the exact same spatial database
as the agent simulation
Figure 3: Early screenshot of 3D rendering application, showing agents depicted as grey boxes in the middle ground. Basic elements are in place, but not a lot of effort has been given to detailed modeling.
Virtual Alpine Landscapes and Autonomous Agents 7
After extensive evaluation, it was decided to develop a custom viewer based on the Openscenegraph library and to develop a custom spatial database format based on the extensible markup language (XML).
The Openscenegraph is a high level scene-graph library, similar in functionality to SGI’s OpenGL Performer, but available for free under the terms of the LGPL (GNU Library General Public License.) It provides a modern high performance real-time viewer that runs on multiple platforms. Plugins exist for importing data from most standard 3D formats. It is also extremely flexible and, being open source, encourages users to customize it extensively.
While the viewing library provides support for many current techniques for real-time rendering (such as view frustrum culling and level of detail management (MÖLLER AND
HAINES 1999)), without data formatted to take advantage of these techniques, the rendering engine is unable to use them. At the same time, the large size of the study area (over 70km2) and the visual realism that we wish to achieve make building the model by hand impractical. As a result of these two factors, considerable efforts are being made to automate the creation of the visual model. A series of software modules are being created that transform GIS data (such as road/path locations, forest types and building locations) into optimised 3D data structures. 3D elements are either directly created by the software modules or linked in from other modeling packages (such as 3D Studio.)
As an intermediate step, the data is exported from the GIS into an XML format, which is the format that both the autonomous agent system and the visualisation system use to build the spatial database. This allows one to quickly introduce new elements into the visual scene (by modifying the underlying GIS) and ensures that any updates will be reflected in both simulations.
6 Next Steps and Conclusion
The basic elements of the system are now in place: the agent simulation is functional, and communicates with the spatial database to understand its environment. At present, the agents only react to objects in their immediate vicinity (~ 5m), and only from the perspective of avoiding collisions with them. The visualisation system works, and displays a highly simplified version of the environment (see figure 3).
The next stages of the project are considerably more difficult: creating the algorithms and model parameters for the simulation. This is made particularly difficult because the route choices depend on a number of inter-related variables such as distance and difficulty in addition to the already difficult visual quality. Increasing detail will be added to the visual model as the agent model is expanded to interpret the corresponding spatial detail.
D. Cavens, E. Lange and W. Schmid 8
7 Bibliography
Bishop, I. D., Ye, W.-S., & KARADAGLIS, C. (2001): Experiential approaches to
perception response in virtual worlds. Landscape and Urban Planning, 54,119-127. Dunn, M. C. (1976): Landscape with Photographs: Testing the Preference Approach to
Landscape Evaluation. Journal of Environmental Management 4, 15-26. Cavens, D. and Lange, E. (2003): Hiking in Real and Virtual Worlds. in Koll-
Schretzenmayr et al. (Eds.) The Real and the Virtual World of Planning. (publication forthcoming)
Gimblett, R. (Editor) (2002): Integrating Geographic Information Systems and Agent-
Based Modeling Techniques. Oxford University Press, Oxford. Hehl-Lange, S. (2001): Structural elements of the visual landscape and their ecological
functions. Landscape and Urban Planning, Vol. 54, 105-114. Lange, E. (2001): The limits of realism: perceptions of virtual landscapes. Landscape and
Urban Planning 54, 163-182. Meitner, M., & Daniel, T. (1997). The effects of animation and interactivity on the validity
of human responses to forest data visualizations. In B. Orland (Ed.) Proceedings of Data Visualization '97, St. Louis, MO.
Möller, T. and Haines, E. (1999): Real-Time Rendering. AK Peters, Natick, Massachusetts.
Müller, H. and Landes, A. (2001): Standortbestimmung Destination Gstaad-Saanenland :Gästebefragung Sommer 2001 Schlussbericht. Bern: Forschungsinstitut für Freizeit und Tourismus (fif) der Universität Bern.
Nagel, K., Beckman, R. J., and Barrett, C. L.(1998): TRANSIMS for transportation
planning. LA- Los Alamos Unclassified Report UR 98-4389. Shafer, E.L. and Brush R.O. (1977): How to measure preferences for photographs of
natural landscapes, Landscape Planning, 4,237-256. Sheppard, S.R. J. (1989): Visual Simulation: a User’s Guide for Architects, Engineers, and
Planners. Van Nostrand Reinhold, New York. Stamps, A. E. (1993): Simulation Effects on Environmental Preference. Journal of
Environmental Management 38, 115-132. Summit, J. and Sommer, R. (1999): Further Studies of Preferred Tree Shapes.
Environment and Behavious 31, 550-576. Wegener, M. and Fürst, F. (1999): Land-use transport interaction: State-of-the-art.
Berichte, No. 46, Institut für Raumplanung, University of Dortmund.
Mixed Realities: Improving the Planning Process by Using
Augmented and Virtual Reality
Joachim KIEFERLE and Uwe WÖSSNER
1 Introduction
Over the past years computer based spatial representations have evolved rapidly from non-real time to real time. Whereas in former times the results of rendering computations could mainly be reviewed with a delay of minutes or even hours, today the planners can interact directly with their virtual three dimensional models like parks, buildings or urban design. This supports planners like landscape architects, architects or engineers in computing, evaluating and communicating their designs more easily. By anticipating the visual representation of a planned project, it enables even strangers in a subject to judge a project more easily and furthermore enables the specialists to optimize a project.
This process can be supported by showing non-visual properties, by coupling the visual information with interactive online simulation. It is one of the main advantages of the artificial realities, that compared to "real" reality several realities can be layered. By coupling the artificial realities, like augmented reality (AR) with virtual reality (VR), further benefits could be observed. This paper will give an overview over different approaches and experiences of the authors. These approaches keep the architects way of working in mind, that many factors like design, structural engineering or environmental effects have to be understood, valued, weighed up against each other and finally combined into one project.
2 Planning Process
The planning process as a complex process with many chances for misunderstandings (SCHÖNWANDT, 1986) can be described in a simple abstraction as a process of ordering, valuing and communicating. All representations supporting the actors in acquiring the necessary information to find the "best solution" should be used. Some examples are:
• visible information, the visual appearance of the planned project • properties of the project / elements • impacts of the project / elements
Virtual representations provide a chance to not only show the visible, but to show the invisible, even ideas and concepts.
J. Kieferle and U. Wössner 2
Fig. 1: The meaning triangle (SOWA, 2000a; modified)
Semiotics, the science of signs, describes the relation between “concept”, the idea in our mind, “object”, the physically existing and “symbol or language”, the notation representing the idea and the physically existing (see fig. 1). By "shifting up" the level of representation, from real to virtual using a notation of a higher level concept, a clearer basis for communication can be achieved.
In this way virtual realities provides a chance to widen the representation notations and thus the way of communicating as well as reducing possibilities of misunderstanding. Especially combining "real" (visible) and virtual representations at the same time proved to work very well. A good example of virtual representation is simulation.
3 Scientific Online Simulation in Architecture
Online simulation with its immediate feedback assists users in understanding complex relations. By changing and adjusting different parameters and observing their effects, projects can be optimized in a very short time. Two examples shall show some possibilities.
3.1 Example 1: Thermal Simulation
With an online thermal radiation simulation, a complete process chain has been implemented from CAD over the simulation to visualization in a virtual environment. The diagram (fig. 2) shows the utilized components. The simulation (Sunface) communicates with the CAD system (AutoCAD) by means of an OLE/COM interface. The geometry might also be read in from a DXF file. Within the simulation, the polygonal data is further tessellated if necessary to better discretize the dataset for the numerical simulation, which then computes the different radiation intensities in the building or structure, taking into account the time and position on earth, material properties, shadows and more. These results are then transferred to the visualization system COVISE, where it is processed by different modules and finally displayed in the virtual environment. Within the virtual
Mixed Realities: Improving the Planning Process by Using Augmented and Virtual Reality 3
environment, different parts of the geometry can be picked, moved and scaled interactively. After doing modifications, the changed geometry is transferred back to the simulation, which immediately computes a new solution. The results are again presented to the user in the virtual environment so that he can immediately verify his modification. Computation time can vary depending on the complexity of the geometry from one to several seconds. It should not exceed a maximum of approximately one minute as no user would want to wait longer for a result.
Fig. 2: Online simulation with feedback loops to post processing, simulation and CAD
3.2 Example 2: Wind Simulation
Fig 3. Simulation of the current of air in an urban / skyscraper situation
To understand the wind impact of for example buildings, computational fluid dynamics (CFD) are used to calculate the airflow. With StarCD or Fluent, simulations can be
J. Kieferle and U. Wössner 4
calculated and parameters can be changed directly from within a virtual environment. The CFD code of these two common commercial products is tightly integrated into COVISE. Streamlines, colored cutting planes, iso surfaces and other flow visualization methods can be displayed together with architectural models (see fig. 3). By placing streamlines at different positions and observing the effects, an understanding of wind behavior can be achieved.
4 Project Reviews Using Virtual Reality
Traditional representation techniques for architecture are drawings, models and renderings. These are representations that either differ in scale or dimension from the final project, that demand the observers’ imagination to transfer what can be seen to what will be built. Especially drawings, due to the necessary abstraction when representing a three dimensional project on a two dimensional medium are a base for many misunderstandings. By using VR in a CAVE environment (CRUZ-NEIRA ET AL. 1993) (see fig. 4), the project participants can physically stand in a scene and get a feeling of being there. The CAVE, a cubic projection room with a side length of approximately 2.5 meters provides space for discussions with up to seven participants.
Fig. 4: Scheme of a CAVE
Mixed Realities: Improving the Planning Process by Using Augmented and Virtual Reality 5
Compared to presentations with a monitor or a head mounted display, the group discussion aspect is one of the main advantages. It could always be observed, that within minutes after the beginning of a session, the participants focused on the scene content and forgot about the hardware involved. Although the stereoscopic images just fit for the person with the tracked glasses, the impression for the other participants with non tracked glasses is still good.
A six degrees of freedom mouse with three buttons is used as the main input device. It showed itself to be a very flexible input device, as it can be easily handed from person to person to explore a scene.
While interaction methods such as movement known from real world interactions are very intuitive, interaction methods not well known from the real world such as scaling, clipping or flying need some experience. For architecture presentations special interaction methods and procedures had to be implemented such as:
• walk mode: while moving around, users stay on the ground, even on a virtual site or stairs.
• object move: single objects or groups of objects can be moved, rotated and scaled. • color picker: elements color is changed with an rgb-cube interface • material changer: material is changed interactively • exhibition picker: single exhibition elements out of a case can be picked and situated
in the scene
5 Project Reviews Using Augmented Reality
Whereas project reviews with VR are well known, there are not that many projects applying AR (some samples see e.g. http://www.hitl.washington.edu). These project reviews profit from the overlaying of the virtual world over the real world, from Virtual Reality over reality. AR can be used both for model reviews (scaled) as well as for real site (1:1) reviews.
5.1 Interfaces and Tracking
Two basic questions arise from applying AR:
1. What is the appropriate interface for flexible working on-site? 2. How can the positions be tracked, to match the virtual world and reality in orientation and scale?
For the Augmented Reality applications, the authors use two experimental setups:
• A head mounted display, the Cy-Visor with attached PAL-cameras (see fig. 5). • A consumer DV camera connected via firewire and a standard monitor.
J. Kieferle and U. Wössner 6
Fig. 5: Cy-Visor with attached PAL cameras.
Whereas the Cy-Visor is intended mainly for single users, groups can review projects on a monitor. Especially for short distances up to a few meters, the Cy-Visor supports the stereoscopic impression by providing different images for left and right eye.
5.2 Application
Video capture and marker recognition is currently based on ARToolKit (BILLINGHURST, M. AND KATO, H., 1999). Markers positioned in the real world (see fig. 8a) or model (see fig. 8b) provide the necessary visual information to match virtual and real images.
ARToolKit has been seamlessly integrated into the VR system COVER, the renderer of COVISE. Marker recognition is done in the capture processes by ARToolKit and the transformation matrices are then used to either move objects or adjust the viewpoint. To allow easy integration in existing Visualizations and rapid development of new applications, a special AR sensor Node was added to the VRML97 support. It acts similar to traditional Plane or Cylinder Sensors. By using this node, the position and orientation of a marker in space can be used to move and orient a VRML object accordingly. Another special VRML group node can be used to render invisible objects, this allows to model part of the physical objects to serve as occluders for the virtual objects. Otherwise virtual objects might occlude real objects in the video picture.
5.3 Examples
AR as a stand-alone application was successfully used for:
• Overlaying architecture models or real world with scientific simulation results. • Switching through various digital model alternatives (see fig. 6).
Mixed Realities: Improving the Planning Process by Using Augmented and Virtual Reality 7
Fig. 6: Virtual model in real architecture model scale 1:500 - a) model with marker b) variant 1 with marker c) variant 2 occluding marker
As the interaction with a real architecture model is very common to the users, they can interact easily from the very beginning. A difficulty in handling these models is that the markers have to be kept clearly visible to the camera at all times. Another drawback is the moderate resolution of standard video cameras (720 x 576 pixels) and the small field of view of head-mounted displays. Fiducial markers which are used by ARToolKit are easy to handle, they can be printed out in different sizes and can be attached to any object. One issue, though, is the low accuracy, especially in outdoor applications. Filtering methods can be used for object tracking as described in section 6.2 but for viewpoint tracking, filters would impose additional lag which leads to deviation as well. Other vision based tracking systems would have to be used to substantially improve registration accuracy.
Fig. 7: Setup of AR on site
J. Kieferle and U. Wössner 8
Fig. 8: a) Image like recorded from the camera with large visible marker b) Same image in AR overlaid by virtual landscape
6 Coupling Virtual Environments
6.1 Coupling VR with VR
By coupling two or more virtual environments (such as CAVEs), the group collaboration of collocated groups is extended to world wide team collaboration. Collaborative sessions can be initiated through a web interface. It provides different rooms where meetings can be scheduled. A central server (VR-broker, VRB) is assigned to each room which distributes messages between the virtual environments and provides a session management. Video and audio conferencing tools can also be started through the web interface.
To coordinate concurrent interaction in a virtual environment, three different collaboration modes were implemented (WOESSNER, U ET AL. 2002):
• Loose Coupling: All partners can navigate through the virtual world independent of each other, they are represented as avatars in the other’s worlds.
• Tight Coupling: Viewpoint and scale are synchronized between all participants. Whenever one of the collaborators moves the world, it moves the same way at all sites.
• Master/Slave: Same as Tight Coupling mode but only one of the participants can navigate through and interact with the virtual world.
The Loose Coupling mode is especially suited for architectural walk throughs where the partners explore a building or landscape at scale 1:1, whereas the tight coupling modes are good for joint work on a model scale building. Master/Slave mode is best used for presentations because undesired interference from other participants is avoided.
6.2 Coupling AR with VR
Augmented reality techniques can not only be used to overlay virtual over real objects but they can also be used to implement tangible user interfaces for 3D applications (WELLER, P. 1993), (FRITZMAURICE, W. ET AL. 1997). Markers attached to physical interaction
Mixed Realities: Improving the Planning Process by Using Augmented and Virtual Reality 9
devices are tracked, using the same computer vision techniques as in traditional augmented reality. In a research project (in DaimlerChrysler MarkenStudio), several toy cars and further elements are equipped with different markers. A camera is positioned above a physical model which defines the interaction area. The VRML model of the same part of a building was created which links the markers to virtual cars and thus the virtual cars can be positioned by moving the physical cars (see fig. 9). The AR sensor has been extended to support collaborative work which means that in a collaborative session, all instances of a virtual object are moved the same way as in the AR setup. This allows linking up the physical model with an immersive virtual environment like a CAVE through a collaborative session. While the cars or furniture in the physical mockup are moved, the 1:1 scaled situation in the CAVE (see fig. 10) is updated continuously to reflect the changes. So the arrangement can be judged in 1:1 scale.
Fig. 9: AR setup for tangible interfaces to position objects in a virtual environment
J. Kieferle and U. Wössner 10
Fig. 10: Exhibition area at scale 1:1 in the CUBE, a CAVE-like environment.
7 Conclusion and Outlook
It has been proven in several projects that the different approaches work well. However each technology and combination thereof has its own benefits and limitations. Whereas VR as a complete virtual environment allows more unbound representations, AR as a context based representation referring to the real world should be limited to representations fitting into the real world appearance (e.g. in scale and orientation).
The group discussion aspect is best supported in a CAVE-like VR environment. Due to the absence of any distracting factors, the participants focus on the planning topics and immerse into a scene very fast. The immersion can be supported tremendously by using accompanying auditory information.
Depending on the interface and application, the AR approach is more focused on a single person experience. The main effect is that in the real world with all its sensory information the information of planned projects overlaps reality. Combining AR with VR, using AR as tangible interface, gives a real time feedback of a model scaled world in a 1:1 dimension.
Mixed Realities: Improving the Planning Process by Using Augmented and Virtual Reality 11
VR as well as AR are still emergent technologies. Whereas VR is pretty common in productive fields, AR is more in its beginnings and there's still a lot of work to be done in the projection interfaces (like resolution in displays, glasses) as well as the tracking. This will be part of a currently planned research project, which should also evaluate the authors observations.
By representing the future impression of a project and combining it with further information, both technologies can improve the planning process and assist planners and laymen in achieving better design.
References
Behringer, Reinhold; Klinker, Gudrun; Mizell, David (eds.): Augmented Reality: Placing
artificial objects in real scenes. IWAR 1998, San Francisco, 1998. Billinghurst, M. and Kato, H. (1999). Collaborative Mixed Reality. In Proceedings of
International Symposium on Mixed Reality (ISMR '99). Mixed Reality--Merging Real and Virtual Worlds, pp. 261-284.
Cruz-Neira, Carolina; Sandin, Daniel J.; DeFanti, Thomas A. (1993): Surround-Screen
Projection-Based Virtual Reality: The Design and Implementation of the CAVE. Computer Graphics Proceedings, Annual Conference Series.
Fritzmaurice, G.; Buxton, W. (1997): An Empirical Evaluation of Graspable User
Interfaces: Towards Specialized, Space-Multiplexed Input, Proc. CHI97, ACM Press, New York, 1997, pp. 43-50.
Kieferle, J. and Woessner, U. (2001): Showing the Invisible: Seven Rules for a new
Approach of using immersive Virtual Reality in Architecture. Proceedings eCAADe 19, pp 376-381.
Schönwandt, Walter (1986): Denkfallen beim Planen. Bauwelt-Fundamente; 74. Vieweg, Braunschweig, Wiesbaden.
Sowa, John F.: (2000), Knowledge Representation: logical, philosophical, and
computational foundations, Pacific Grove, Calif., Brooks/Cole, pp. 192-194. Weller, P. 1993: Interaction with Paper on the Digital Desk, Comm. ACM, vol. 36, no. 7,
pp. 87-96. Wössner, U., Schulze, J. P., Walz, S. P., and Lang, U. (2002): Evaluation of a
Collaborative Volume Rendering Application in a Distributed Virtual Environment. Proceedings of the 8th Eurographics Workshop on Virtual Environments (EGVE), May '02, ACM Press. 113-122. 2002.
Designing a Virtual Reality
Nyungar Dreamtime Landscape Narrative
Lorne LEONARD
1 Introduction
This article demonstrates the use of Virtual Reality simulation as a new pedagogical tool
for cultural landscape preservation and education. The case study examines an Australian
Aboriginal landscape conception known as the “Dreamtime”. To many Australian
Aboriginal nations, the Dreamtime is an important religious concept that describes the
meaning and creation of the cosmos; it is a complex system of signification. The prototype
virtual landscape focuses on the Dreamtime narratives from one nation, the Nyungars, who
are the indigenous nation from the southwest of Western Australia. Many forms of
Dreamtime narratives exist, and this project focuses on those that explore the landscape and
its creation. One narrative in particular, that describes the formation of the Gabbee Darbal
(Swan and Canning River) by an ancestral being called the Waugal, helped to inspire the
author to design and code a virtual reality landscape within the Cave Automatic Virtual
Environment (CAVE™) at the National Center for Supercomputing Applications (NCSA™)
at Urbana Champaign. This Nyungar narrative aided in creating a virtual landscape
narrative that helps to visualize Nyungar sacred landscapes severely affected by
colonization. A crucial principle explored by the author, was how to represent the Nyungar
culture, a culture nearly erased by colonization in a sensitive and accurate way. This
exploration, in conjunction with landscape narrative practices found within Nyungar
narratives, assisted in creating a set of design guidelines for forming the virtual landscape.
Furthermore, the methods of creating the scenes within the virtual landscape are explained.
2 Who are the Nyungars?
The Nyungars are the Aboriginal nation located in the southwest of Western Australia. In
an attempt to understand Nyungar concerns, perceptions and knowledge of the southwest
area, the author selected a specific landscape region known by the Nyungars as the Gabbee
Darbal and focused on the Whadjuk sub-nation area. Located centrally within these
landscapes is the city of Perth and its surrounding suburbs, where approximately 2 million
people currently live. Colonization has destroyed many Nyungar sacred sites/landscapes
and only a few survive. For many years, Nyungars and others have been fighting for
acknowledgement by the Australian government and the public concerning their cultural
heritage. The purpose of this project is to visualize a specific aspect of Nyungar culture,
Nyungar Dreamtime landscape narratives, in particular, the formation of the Gabbee Darbal
river landscapes. Many of these concepts are physically impossible to create and difficult to
explain without being immersed. To re-create these landscapes and concepts, the author has
created a computer program and designed a virtual landscape that allows users to
L. Leonard 2
experience immersion using the CAVE™. The virtual landscape design is not a replication
of the contemporary or past Gabbee Darbal landscapes but rather, a design that encourages
users to conceptualize these Dreamtime landscapes.
To discuss and represent Aboriginal issues and concerns as a non-Aborigine inherently
raises many complex issues, and to address all of these is beyond the scope of this paper.
To represent Nyungar culture, it is critical for one to challenge prejudices and stereotyped
beliefs. Ideally, the approach of this project would be to work with the Nyungar
community, to make personal connections and exchanges at all stages of this project. Due
to logistical constraints, this project relies on the published information that Nyungar
people have generated and shared with other communities, focusing on narratives that
relate to the Dreamtime. Furthermore, this project examines ways to represent these
narratives in conjunction with the author’s perceptions and experiences of growing up
within the study site region. It explores the interplay of the author’s cultural identity of
what it means to be ‘Australian’, with the Gabbee Darbal landscape being a fluid space for
negotiation. Presently, the stories seen and told amongst the Gabbee Darbal landscapes are
predominantly non-Nyungar. These landscapes have both cultural and natural history
implicit to Nyungars and non-Nyungars, and are continually generating narratives for
further and new types of narratives. Multiple narratives with objective and subjective
elements that interconnect at places throughout the Gabbee Darbal landscape are yet to be
told. These places are continually engaged in symbiotic and ongoing relationships with
people and the landscape. These extend well beyond its specificity and the borders of our
past and present ‘versions of time’. They are Dreamtime narratives.
3 The Dreamtime and Landscape Narratives
Australia’s landscape is ancient and unique. As one walks through such a landscape, one
can start to perceive a rich history and traces of one of the longest living civilizations, the
Australian Aborigines. Australian Aborigines did not build temples, but instead, they
revered the landscape by creating elaborate narratives about their culture and their
connections to the landscape. The Arrernte Aborigines ‘shared’ concepts about their
religious practices and values with anthropologists Spencer and Gillen, who translated
these concepts into the English language and conceived the term “Dreamtime”(Spencer).
Dreamtime reflects Aboriginal cosmologies and religions, and in particular, the creation of
the universe. Also known as “spiritual” or “mythical” beings, the ancestral beings epic
journeys created the universe. Many Aboriginal narratives describe the ancestral beings’
ability to transform many times into other forms, such as humans, animals, or inanimate
objects. A common ancestor amongst Aboriginal nations is the “rainbow serpent”. As the
rainbow serpent and other ancestral beings created the world, the process was ‘mapped’
onto the landscape. The Dreamtime realm is independent of linear time; rather it is another
dimension of reality (Morphy). For many Aborigines, the Dreamtime has never ceased to
exist, with no beginning and no end. The Dreamtime is the basis of the present, and
influences the future. With this belief, the present is as much a feature of the future as it is
of the past. The Dreamtime relates to space and time, referring to the origins and powers
that are located in places and things; they can be described as landscape narratives.
Designing a Virtual Reality Nyungar Dreamtime Landscape Narrative 3
3.1 Landscape Narratives
The Dreamtime connects events, sequences, memory, space, and other abstractions to the
more tactile aspects of place. They order and configure experience of space into significant
relationships, offering ways of knowing and shaping landscapes. To truly experience the
Dreamtime is to transcend boundaries and resonate with other dimensions of experience. In
a sense, the Dreamtime represents stories that all Australians participate in and these stories
shape the Australian landscape. They are a process of remembering and interpreting the
landscape; they give form to space and experience. To participate and follow the
Dreamtime is more than just listening to a story. As it tells of origins, explains causes,
marks the boundaries of what is perceivable and explores the territories beyond (what is
told), it is a narration (Potteiger). Furthermore, the Dreamtime consists of stories expressed
(means of telling) through the landscape, orally and through other forms of art media
(dance, paintings and film). For many generations, Aborigines have practiced mapping
landscapes into the very texture and structure of stories. Many of these Dreamtime
narratives discuss how the ancestors shaped the landscape, and in turn how places
configure narratives, as opposed to considering the landscape as only the background
setting. Thus, in the Dreamtime, the landscape is integral to the narrative and hence they
are landscape narratives (Potteiger).
3.2 Following Nyungar Dreamtime Narratives
To understand these narratives is to read the landscape through association, events and
stories that are part of the physical and non-physical form of the Australian landscape. To
‘read’ the landscape requires guidance. One way to navigate or follow an established
hierarchy within the Dreamtime in a spatial manner is through what some Aboriginal
nations refer to as song cycles or songlines. Songlines weave the ‘dreamer’, through the
landscape following ancestors’ events, enabling the dreamer to recount specific tales,
allegories or social narratives. As Aborigines (re)-create songlines, they place elements to
form sequences, plus they interpret empty spaces and create meanings; thus, Aborigines
become authors. In addition, the landscape informs Aborigines about cultural and natural
processes by recording these changes; in effect, the landscape tells a story. These changes
or effects may or may not be ‘natural’. For example, Nyungars used fire to hunt kangaroos
and have narratives describing such events (Bates). With fire, Nyungars changed the
landscape, creating places that tell new narratives. Often these narratives discuss the
progressive stages of actions or events of nature. Water, being essential for life, dominates
many of these narratives. The Waugal ancestor is the protector of water environments and
there are many Nyungar narratives describing how it creates water features (rivers, streams
and aquifers) that constitute and shape the Gabbee Darbal landscape. For example, the
following describes characteristics of water flow: “Noongar from out around Brookton and
York talk about how the Waakarl came out of the earth. It went different ways, making
tracks through the hilly country. Sometimes it went kardup boodjar (under ground) and
sometimes it went yira boodjar (over ground). The Waakarl’s kaboorl (stomach) pushed the
boodjar (earth) and boya (rocks, stones) into kart (hills). You can see the Waakarl’s path in
the shape of the boodjar (ground / land)” (Collard 2000).
L. Leonard 4
4 Design Strategy
Actions by the Waugal formed the basis to develop a design strategy, however as the
author is not a Nyungar, it became crucial to investigate ways to ‘represent’ Nyungar
culture. An essential and difficult task was to discover personal and others’ perceptions
about Australian people and place, and to question various versions of information
portrayed. It was important to reveal racist discourses and seek answers to questions in an
accurate, non-racist, and sensitive manner from the images and text that ‘represent’
Aboriginal culture. This project represents the ‘final’ stage of many such processes that the
author explored and developed. Briefly, in an attempt to be ‘accurate’, information
represented and used focused on the micro and specific interests of Nyungar culture,
keeping in mind the following limitations: (1) That information is interpretive and is an
artifice. (2) That the designer has a role and impact on guiding users’ perceptions, only
being able to represent information from a selection of Nyungar accounts that is perhaps
similar to other viewpoints. (3) To recall and critique that, at times some resources are the
result of individual subjectivity. (4) To beware of the hazards of the audience perceiving
specific experiences as ‘typical’ of an entire community, or how an individuals’ voice may
be perceived as the voice of the whole community. In this project, Nyungars appear
linguistically as agents in the production of knowledge and as an inspiration for creative
activity and interpretation. This project is part of an ongoing process of understanding the
ways Australian people create culture and history. It derives from and reacts against
historical representations and symbols of Aboriginality: finding new ways to ‘view’ and
understand the Gabbee Darbal landscapes using virtual reality, rather than emphasizing the
differences or the contrasts between people. The project aim is not to construct a
replication of the Gabbee Darbal landscape as it was before colonization or how it is today.
Rather, the aim is to explore the entangled memories of the past, the present and ultimately
the future and their relationship with the inter-subjective state of experience.
The methodology to formulate ideas or guidelines to design the virtual landscape was based
on 5 landscape narrative practices as discussed by Potteiger: Naming, Sequencing,
Revealing and Concealing, Gathering and Opening. The majority of participants using the
CAVE™ version of this project are likely to have limited knowledge about Australian
culture. Thus, it was important to create a strategy to introduce users to this knowledge.
The final design strategy consists of 7 distinct scenes to form the following sequence and
purpose: (1) Introduction to ‘Australia’ (Fig. 1), (2) Introduction to ‘Aboriginality’ (Fig. 2),
(3) Introduction to southwest of Western Australia (Fig. 3), (4) Abstraction of Nyungar