Digital Environments: Design, Heritage and Architecture
Melissa M Terras
Virtual Reality and Archaeological Reconstruction
Keywords: Virtual reality, VRML, 3D modelling, archaeology, HTML
Virtual reality, interactive computer generated sites and scenarios, theoretically creates great opportunities for archaeology, history, and education. Immersive computer driven environments impart information regarding space and human experience that would not be possible using traditional means of representation. Places and structures too remote, dangerous or deteriorated to visit can be experienced, and virtual models can provide the context in which to understand other complex issues surrounding an environment. In the past the technology required to produce such virtual 'worlds' has been expensive and complex, but Internet developments in the last few years have provided the means to generate three-dimensional interactive worlds cheaply and quickly. VRML, Virtual Reality Modeling Language, allows anyone to easily build 3D computer models of objects and places without specialised equipment, and more and more of these virtual archaeological models are becoming available on the Internet. However, few models have been evaluated to assess their educational or archaeological worth, and there remain technical and theoretical issues which need to be addressed regarding the use of VR and VRML in archaeology.
The Sen-nedjem Project, undertaken at the Humanities Advanced Technology and Information Institute, University of Glasgow, Scotland, between May and October 1998, investigated the success of an archaeological virtual reality model for use in a museum context by building an interactive computer model of an Egyptian tomb based on pre-published archaeological evidence and testing this model with a view to installing it in the Kelvingrove Museum, Glasgow. The project illustrates the process of developing such a display and suggests that although it is technically possible to create a virtual reality model based on archaeological evidence there remain various problems regarding the use of VR in archaeology. The scarcity of available spatial information reduces the authenticity of these virtual models, and a lack of meaning and purpose inherent in their medium complicates their educational value. Also, many technical issues still need to be resolved before virtual archaeological reconstructions become commonly and effectively used for pedagogical purposes in archaeology and the humanities.
Although three dimensional representation has been used in archaeology for many years and multimedia is becoming increasingly used in museums, the costs involved in producing a three dimensional model for public use have proven prohibitive until recently1. The swift growth of the Internet and the development of associated software, hardware and protocols has provided a solution2. A three dimensional sister to HTML (Hyper Text Mark-up Language), VRML (Virtual Reality Modeling Language) was conceived in the spring of 1994 at the first World Wide Web conference in Geneva, developed in an on-line forum, and adopted by the International Standards Organisation as the web standard for three dimensional environments in June of 19963. A VRML document, identifiable from the suffix '.wrl', is an ASCII text file that describes a three dimensional scene. A list of commands instructs the computer to place objects of given forms in certain places within the Cartesian grid of the 'world'. Basic objects can be constructed from pre-defined spatial primitives, such as box, sphere, and cone, which generate small, quickly rendered code, and more complex objects can be defined as separate faces and planes. It can take thousands of polygons to create a single building or landscape in the world. Basic animation, lighting, and interactivity are supported, and the world can be linked hyper-textually to other VRML and HTML documents. World files can either be scripted by hand, exported from various CAD programs, or written with the help of a 'world-builder', an application that facilitates the creation, placement and sizing of objects within the scene. A relevant browser plug-in such as Silicon Graphics' CosmoPlayer parses this ASCII text file, and the resulting representation can be viewed and interacted with on the majority of systems available today simply using the mouse and screen. All browsing is done on the client machine resulting in low bandwidth requirements and hardware independent distribution. As the software is readily available and VRML is platform independent, these virtual worlds should be able to be produced relatively cheaply and easily using data from archaeological reports4.
That said, VRML is not the optimum virtual reality system currently available. It is rather a 'jack-of-all-trades', providing many basic functions that are designed to run on all platforms and was never designed to replace more sophisticated VR systems intended for specific tasks or configured to run on specialised hardware. There remain some problems with its execution on different machines and platforms; colours and lighting are not standardised between browsers, and due to the fact that the browsers take mathematical short cuts whilst rendering the world there can be slight differences in a world each time it is viewed. Problems abound for the programmer: VRML is very case and syntax sensitive and errors that may not produce obvious effects in their own routine may result in errors in the remainder of the file, making it hard to debug. VRML world files often run into thousands of lines of code which can be very hard to navigate, even with extensive remarks. Although these points must be taken into account, VRML is a standard, and the best standard that exists at the present time for the construction, publication, distribution, and viewing of virtual worlds on different platforms. However, the future of VRML is currently in question due to the developers of the associated software (Silicon Graphics, then Platinum) ceasing production of any further VRML based systems. Now open source code, VRML may either evolve into a more sophisticated language or cease to be used (see www.vrml.org for updates).
Constructing the Model of Sen-nedjem's Tomb
It became obvious during the project that there would be several clearly defined processes in constructing the model of the tomb; requirements for the model were discussed with the museum, an archaeological site was chosen, archaeological evidence was collated, the basic model was constructed, images were prepared and mapped onto this basic model, the model was refined and animated elements added, the user interface was designed, and testing and development were carried out. Although is it acknowledged that any issues raised here pertain to one particular instance of the construction of an archaeologically based VR model, many points can be taken to be indicative of problems encountered when translating any archaeological evidence into VR.
Defining the Project Requirements
The Egyptian gallery in the Kelvingrove Art Gallery and Museum has remained the same for over thirty years, meaning many visitors pay little attention to the artefacts within this particular area of the Museum. The curator of Ancient Civilisations at the gallery, Simon Eccles, suggested that an injection of energy and life was needed to encourage visitors to spend more time there, hopefully becoming familiar with more of the displays and exhibits as they did so. It was felt that an interactive, educational multimedia electronic display placed in the gallery could offer the solution to this problem. With the growing technological developments in virtual reality (and complementary public interest) a computer model of an Egyptian tomb could be used to give some context to the artefacts in the gallery; persuading the museum visitor to pay more interest to the existing displays.
From the outset, then, the Museum had a firm idea of what they required from the proposed model. However, museum staff had little experience of computing and no experience of VR and so were not aware of any issues and problems which could complicate the construction or display of such a model, and no realistic expectations of what such a model could achieve; a problem of inexperience encountered in numerous humanities projects which attempt to use IT as a solution5. The most fundamental problem which resulted from this inexperience was the museums' insistence that no ancillary information regarding the structure was to be contained within the model itself (ie dates, geographical information, historical information), and that they would provide any such information to be displayed alongside the model. This immediately divorced the model from a range of necessary information which would help the end user to understand the structure: the ramifications of this becoming apparent when the model was subject to user testing at a later stage. Although, because of the time scale in this project, it would not have been possible to integrate such information into the model anyway, it draws attention to the fact that when defining the requirements of such a model great care has to be taken from the outset to ensure all parties are aware of the consequences of such decisions. A certain degree of understanding regarding the use of multi-media in an educational context must be present before any such requirements can be suggested or finalised.
Choosing the Site
The tomb of Sen-nedjem (a workman on the royal tombs in the Valley of the Kings from the 19th Dynasty) at Deir el-Medina was identified by the Museum for possible reconstruction for various reasons. A model of this tomb would reflect the Egyptian collection of the Kelvingrove Art Gallery and Museum, most of which comes from such individual tombs. Also, as Egyptian tombs go, it was thought to have fairly detailed archaeological reports which would hold the information required to reconstruct the site6. A recent publication contained comprehensive full colour images of the tomb7, and, once copyright clearance was obtained, it would be possible to use these to produce a photo-realistic reconstruction of the inner chamber of the tomb. The tomb was also a suitable size for reconstruction, was not too complex, and contained unique, detailed and well preserved wall paintings.
Collation and Problems with Archaeological Evidence
The first major obstacle encountered was the archaeological evidence. As this was to be an educational reconstruction it was imperative that the model was as archaeologically sound as possible, but it soon became apparent that the tomb was not as well documented as first thought. Access to relevant literature was limited, and the reports failed to supply comprehensive measurements of major surfaces. There was little documentation regarding the courtyard and the main entrance of the tomb complex, and the plans of the tomb supplied in the reports were very basic. Data obtained from the reports and plans was used to construct a tentative new set of archaeological drawings from which some missing dimensions could be surmised. These plans were used to construct a basic model in VRML 2.0 using the world builder CosmoWorlds, but any completely missing areas and details had to be based on the hypothetical archaeological drawings and other information regarding similar tombs (for example, it was presumed that the floor of the tomb was of packed earth). Also, due to the fact that the majority of artefacts listed in the archaeological reports are now lost, it was decided that the tomb would have to be presented without any contents. It is acknowledged that the model is flawed due to the quality of data available; a problem which will face others should they try to construct models from the majority of archaeological reports available. This is not necessarily the fault of these reports as such publications are never expected to provide enough complex data to reconstruct a site in three dimensions; spatial data recovered in an excavation is rarely published in its entirety. As modern excavation recording turns towards electronic methods it may become easier to obtain the primary data the publications are based on,8 but in general unless the model builder has access to all archaeological site records they will not have enough information necessary to build an authentic VR model.
Figure 1: The basic model viewed from above before texture mapping.
Image development and texture mapping
One of the features of VRML is that images can be 'texture mapped' onto surfaces, allowing photo-realistic models to be created. Small images are tiled over a larger area to give a repetitive surface, reducing file size and rendering time. Making the model photo-realistic involved two procedures; producing small images for the application of textures such as stone, sand, and grit, and constructing larger images to make the individual faces of the painted barrel vaulted chamber. Smaller texture images were created by scanning pictures of the site as it is today, and taking samples of the dirt, sand, and stone found there.
Figure 2: A view of the tomb complex from inside the courtyard. This shows the use of the image textures to give the background and courtyard ground their texture.
Preparation of the larger images was more complex. As each wall surface in the inner tomb was covered with individual wall paintings the tiling mechanism was not applicable and images had to be prepared which would cover each wall surface fully. The first issue confronted was the resolution and size of the images needed to cover such a large virtual area. VRML browsers allow the user to 'zoom in' to surfaces and, because each wall painting contained so much detail, it was important that this manoeuvre did not produce a pixelated unintelligible mass of colour when the surfaces were viewed at close range. It was discovered that the minimum resolution needed for the wall surfaces was 150 points per inch, resulting in a file size of over 2 MB for each wall surface, and although various attempts were made to decrease the file size it was found that these large files were required to ensure the visibility of the tomb paintings in the computer model.
The mapping of the images onto the virtual tomb surface was not a straightforward process. Many of the walls were photographed in two or more segments, and the roof was recorded by a series of ten images, meaning a composite image for each surface had to be constructed. Once again, it was also discovered that many surfaces on the tomb had not been properly documented. Panels and inscriptions were missing from the "comprehensive" photographs, particularly the large panel containing hieroglyphs depicting the Book of the Dead beneath the image of Anubis and the mummy. It was decided to try and recreate such areas to give the overall feeling of the painted vaulted chamber, but it was noted that the authenticity of the model was fast decreasing.
Figure 3. The reconstructed hieroglyphs beneath Anubis and the mummy. Although some repetition is evident in the hieroglyphs, it was considered that it was necessary to try and reconstruct the panel to blend with the rest of the model rather than leaving it bare. The hieroglyphs were taken from other parts of the frieze in the tomb.
These completed images were then mapped onto the surfaces of the tomb. This required much trial and error, but the final result meant that a more realistic appearing model was created. However, the major drawback with using such large images was the time taken to render the model. Once loaded, the browser copes admirably with the large images, allowing smooth movement within the tomb. Unfortunately the model takes approximately five minutes to render (running from the hard drive of a 330 MHz PII with 64 MB RAM). A further attempt was made at file size reduction, but this caused great distortion to the images.
Figure 4: A view of the inner chamber of the tomb model. This illustrates the left wall, the barrel vaulted ceiling and the lower long wall. The viewer is standing just in front of the entrance to the chamber. The two large figures illustrated on the large left wall are that of Sen-nedjem and his wife Eineferti.
Refining the model
After the basic construction of the model, it was then decided to add other features that would improve the presentation and user experience. Interactivity was added with a moving door and tomb covering slabs. Lighting was experimented with. A series of viewpoints were defined, and a seventy second animation of the model was constructed. A sky and sun were added, with the external lighting adjusted to make the experience more realistic. A landscape was added to place the model in some context. An attempt was made at constructing images of the Deir el-Medina tomb complex to place the model in the correct setting, but this was found to slow the rendering of the model down a great deal, and so the detailed background was abandoned for the time being (even though this makes the model even less accurate). Other issues considered were the use of sound to create atmosphere, the use of information boxes to explain to the user points of interest about the tomb, and the possibility of making a voice-over to explain what the user was experiencing throughout the animated flythrough.
Figure Five: the final model in its desert setting.
The User Interface
There were many things to consider when designing the user interface. Essentially, this was to be a stand-alone electronic display for use in a museum. This necessitated that the interface be intuitive, suitable for users of different ages, easy to use by those with no technical experience, and interesting enough so that those with some knowledge of virtual reality did not feel that they were being patronized. The design of the interface was limited by the browser plug-in used to view the model, CosmoPlayer 2.0, which had to be integrated with some user information to allow any user to view the model.
CosmoPlayer has its own set of predefined tools which presents a confusing array of options to the user, and only has a basic help mechanism integrated into the system. As many of the advanced options were disabled in the model as possible to provide a basic set of navigational tools; go, tilt, slide and seek, and other steps were taken such as implementing the "collision" tool to disable the avatar's ability to walk through walls. Even with these measures initial testing showed that users were unable to navigate the model themselves due to the complex tools provided in CosmoPlayer, and that the flythrough was the only option where the user could find the main area of the tomb.
To get a user to implement the flythrough required some form of explanation. A frame mechanism was set up which presented the user with an opening screen. The next screen provided information about how to use the model, and the following one implemented the model. There was always a button present which reset the model for the next user. The obvious problem with this was the time the model took to render every time the model was loaded. Also, the design of the system already presumes that the user is au fait with hyperlinks, the mouse, and the concept of returning 'home' to the start. It was obvious that this was not a suitable end product.
The model was evaluated to see if it succeeded technically, archaeologically, and educationally. There are many problems in testing such a display, mostly due to the fact that very little research into the evaluation of such systems has been carried out, affording little framework to base a comprehensive testing strategy upon9. Also, the testing of such a device requires much observation of the user, ideally in the environment in which the final system is to be placed. Care has to be taken that test volunteers have a variety of computing experience to account for the fact that a large proportion of museum visitors will have little or no computing knowledge. The person collecting the data has to as unobtrusive as possible to account for the fact that museum visitors do not often operate or understand an exhibit in the way those responsible for developing the display wanted them to. To try and incorporate all of the above, a program of formative evaluation was adopted to systematically observe a series of potential museum visitors10. Think aloud protocols, structured questionnaires, and informal interviews were all used to question a variety of potential users. Data was collated and any issues raised were clarified through discussion, bringing some empirical evidence to the testing process.
Technically, the model was shown to have succeeded. Although there are a few small areas that could be adjusted within the model it is fully functioning and is in a finished state. However, there were many problems presented by the means which the user could explore the structure. Although an elementary user interface had been designed, the model was tied to using CosmoPlayer 2.1 as its browser plug-in. The toolbar provided by CosmoPlayer was found to be wholly unsatisfactory for users both familiar and unfamiliar with virtual reality, and it was noted in all cases that the experience of the model was greatly effected by the poor tools provided. It is very doubtful that a non-computer literate user would have the required skills to explore and view the model, posing a problem for the display of such a model in a museum. A complete redesign of the user interface would have to be undertaken before the model was used in a museum context, and unfortunately there was not time to do this within the scope of this project.
Although a successful, working computer model, it was obvious that the archaeological evidence did not provide all the information necessary to create an authentic model. Throughout the construction of the model care was taken to document any questionable areas but the relationship between the recorded data and the interpretations made around it is not clear within the final model. It is accepted that such reconstructions remain only one permutation of the possible evidence, but, unlike illustrations, the photo-realistic qualities of virtual reality add some faux authenticity to flawed representation. An uncritical viewer may be convinced that that was how the structure was, not how it could have been; the model may look in places realistic, but it does not express the theoretical nature of its interpreted form. How much of a problem this actually is is debatable: limited work has been done to date on user perception of virtual historical environments, but the fact that this model was due to be used in an educational context suggests the link between VR and actual historical representation is problematic, and heightened by the photo-realistic nature of the medium.
The problem of representation in this particular model is coupled with the fact that although the model may look good it provides no knowledge as to what it is, what the user has to discover, and what is the point of trying to navigate around the virtual space; a common problem with VRML models and one which was imposed on this model from the outset (see above). Some kind of feedback system needs to be incorporated into the model to tell the viewer what they are looking at, what period of history it belongs to, and how the structure relates to the archaeological site as it can be found in the present. Many users approaching the model asked 'Is that it?', 'Have I completed the task?' and 'What is it I'm looking for?', stressing the fact that virtual reality by itself is not the best means of imparting knowledge regarding a structure. Undoubtedly, virtual reality provides an alternative means for viewers to gain spatial awareness of a site (in the case of Sen-nedjem's tomb it showed the complex relationship of one wall painting to another which was not apparent by looking at two dimensional prints in the publications) but without any ancillary documentation the meaning of certain elements will always be beyond the educational limit of a virtual model. It was very interesting to see the wall paintings in situ, but there was no way to explain their content and how this related to the other paintings surrounding them.
There is no doubt that virtual reality does provide an alternative way to view complex three dimensional data, providing an alternative medium for educational and archaeological reconstructions. However, it remains only one means of presenting information and should be used in conjunction with others to ensure the user understands the context and meaning of the computer model. A great deal of thought must be put into what is hoped to be achieved by constructing such a model, and the design of the user interface must be carefully constructed to allow non computer literate users to access the model. It is doubtful whether the tools provided at present are adequate for uninitiated users to view a VRML model.
The Sen-nedjem Project has shown that it is possible to create a virtual display from pre-published archaeological evidence (although older archaeological reports rarely contain the level of detail required to build such a model, and there will be some elements which have to be constructed by the model maker). However, these models carry with them theoretical problems regarding authenticity, and the user must be made aware that any computer model is merely one possible illustration of the past (this may be understood by archaeologists but it is not obvious to a member of the public). There are also logistical problems involved in the creation of such a virtual display, which may still prove prohibitive for an institution. Even with the new technology available it is still a time and labour consuming process, and it remains technically intensive. Costs of software, hardware and staffing remain to be met, and as there are no methodologies in place for the evaluation and testing of such a model, the educational success of virtual reality remains uncertain. VRML does provide the means to create an archaeologically sound virtual reality display, but at present there aren't the user tools available nor the understanding of how it relates to other resources to make it a viable educational tool. Only after further research, experimentation and testing should archaeological virtual models become a common fixture in public institutions.
Dr Seamus Ross, HATII, University of Glasgow.
J. Henderson, University of Oxford.
Verlag Philipp Von Zabern for granting permission to use images from Das Grab Des Sennedjem in both this project and this publication.
1 Economou, M. (1997) 'The Application of Interactive Multimedia in Museums and Archaeology: Design and Evaluation of the Eusperides Program', Doctoral Thesis, University of Oxford.
2 Mitchell, W. L. and Economou, D. (1998) 'The Internet and Virtual reality in Heritage Education', Proceedings of 26th Computer Applications in Archaeology, Barcelona, pp.74-75.
3 Marrin, C., and Campbell, B. (1997), VRML 2 in 21 Days, Sams net.
4 Gillings M., and Goodrick, G.T. (1996), Sensous and Reflexive GIS: exploring visualisations and VRML
5 Economou, M. (1997), 'The Application of Interactive Multimedia in Museums and Archaeology: Design and Evaluation of the Eusperides Program', Doctoral Thesis, University of Oxford.
6 Bruyere, B. (1959), La Tombe No1 de Sen-nedjem, L'Institut Francais - Annales du Service des Antiquities de l'Egypte, Tome XX.
7 Shedid, A.G. (1994), Das Grab des Sennedjem, Verlag Philipp Von Zabern.
8 Burgess C., Gilmour S., and Henderson J.C. (1996) 'Recording Archaeological sites using GIS', Archaeological Computing Newsletter No 46 p.229.
9 Mitchell, W.L. and Economou D. (1998) 'The Internet and Virtual reality in Heritage Education', pp.74-75, Proceedings of 26th Computer Applications in Archaeology Conference, Barcelona.
10 Terras M.M. (1999), 'A Virtual Tomb for Kelvingrove, Virtual Reality, Archaeology and Education', Internet Archaeology, (http://intarch.ac.uk/journal/issue7/terras_toc.html).