Joachim Sauter (Speech at the Doors of Perception 3 Conference)

Table of Contents:
Summary
Introduction
The Terravision Project
The Representation of Space
Potsdammer Platz: A VR System
Telecommunication and Networking
Terravision's Rendering of the Earth
Mirror

* * *
Summary
Joachim Sauter of Art + Com in Berlin devotes his speech to the Terravision project, a system that uses a virtual representation of the earth as an interface for network information. This virtual earth is generated from satellite images and altitude data stored on servers located around the world, integrated into a television system by an ATM broadband network. Users can move about the earth, zooming in from the level of the upper atmosphere to virtual representations of city squares or even individual buildings. Information can be accessed topographically and chronologically: a user can see weather patterns or, for example, zoom in on the Potzdammer Platz in Berlin and use a camera icon (placed at the actual location of the original camera) to view a film of it made in 1929. Sauter explains two paradigm shifts underlying the creation of this system that allows users to `fly' about the earth: the shift away from classic, central perspective; and the shift from remote computing to decentralised network information. The freedom to choose a viewpoint of one's own within a spatial representation confronts Art + Com with a central question: To what extent can this new basis be used for the organisation and communication of information? Terravision is the embodiment of Art + Com's answer, based on a triad of concepts: the fact that we are spatially socialised; the existence of long-standing cultural traditions of using space to memorise information (the Greek art of mnemonics, for example); and third, the idea that exploration from one's own viewpoint always creates superior understanding. Terravision was conceived as a tool to generate knowledge about the condition and future development of our planet.

* * *
Introduction
Art + Com established itself in Berlin in 1988 as a place where designers, computer experts, architects, scientists, artists and engineers can come together to realise their ideas and goals. The method of working at Art + Com is characterised by an openness to new approach and innovation -- a readiness to question patterns of thinking and the ability to speed up complex development processes. Art + Com was founded as non-profit organisation with the intention to promote co-operation between the arts and sciences. In the middle of this year, we founded a profit-oriented company in addition to the original, non-profit organisation. Our main activity in both parts is research into network, virtual reality application and Terravision, which will be the focus of this presentation.

In this project, we use a virtual representation of the earth as an interface for network information. Here, the virtual earth is generated out of satellite images and altitude data, stored around the world, which can be integrated into a television system by an ATM broadband network.

* * *
The Terravision Project
Our motivation for this project was not to create a photo-realistic, virtual representation of the earth. From the beginning, the main goal was to initialise and establish a system which allows users to collect, store, process, exchange and examine any kind of data related to geographic origins. They include ecological information, global topography, satellite imagery, environmental monitoring and forecasting and any supplementary data, such as statistics.

The system is based on two paradigmatic changes concerning space, which we have experienced very practically through our work of the last seven years. The first is the paradigmatic shift in spatial presentation from a single steady view point in classic central perspective to the present ability to interactively select your own viewpoint in a virtual space. The second is the shift in telecommunication and networking from remote computing to the organisation of decentralised network information.

* * *
The Representation of Space
In the renaissance, the changes in the sciences and the arts meant that mathematics had a strong influence on the evolution of painting. This was manifest in the development of central perspective. Here we find one steady viewpoint in the picture. In the last thirty years, we have been able to observe again the emergence of a new spatial representation technique. With real-time computing, we no longer have a fixed viewpoint outside the representation. Instead, the observer can interactively select any viewpoint within the spatial representation. This development confronted us with one main question at the beginning of this decade: To what extent can this basis be used for the organisation and communication of information?

The attempt to organise information in space was motivated by three facts. Firstly, we are all spatially socialised and therefore capable of navigating in 3D rather than in 2D. Secondly, this approach stands in a long tradition that can be based on various cultures. One example is the mnemonic technique of the ancient Greeks, in which they use mental spaces to memorise information. The third reason for organising information in space is that through real-time computing, viewers are now able to choose their own viewpoint in relation to information. And an exploration from one's own viewpoint always leads to better understanding and deeper knowledge.

* * *
Potsdammer Platz: A VR System
I will briefly cover some short excerpts from our former projects that deal with this topic.

In 1992, our architects applied to take part in planning competition for the Potsdammer Platz area in Berlin. Alongside their model proposals, we started to develop a VR study information system, which allows the user not only to navigate throughout city space, but also through time in order to explore spatial and historical information about Berlin.

The user can interactively move through the centre, viewing, for example, the Palace of Republic as it is today, the Old Imperial Palace before the Second World War and the New Parliament Building in the year 2002. Users can move back through the year 1900, where you can interactively look around the centre at its condition at that time.

Motion through time enables the organisation of historical information at the place and time of its occurrence. If the user moves through the square in front of the Old Imperium Palace, he might see an active element which would trigger a film that was taken at this site in 1900. In other words, information is organised and can be explored according to its origin in time and space.

Another example out of the very early days of Art+Com, which dealt with this topic, is the project Home of the Brains, in which abstract ideas are organised in a virtual space. Here the concept of the museum as a public place for storing and discussing cultural information enters into the realm of the virtual. Visitors enter a space and find themselves in a linear dialogue conducted by four media philosophers and computer scientists. Each corner of the space was fitted out with interactive objects for one thinker. Users can intervene in the discussion by interacting with this object, thus defining their own position with respect to the space and the debate itself.

* * *
Telecommunication and Networking
Let me leave the paradigmatic shift in spatial representation to focus briefly on the shift in telecommunication and networking I mentioned earlier.

Deutsche Telecom established its first broadband network in the mid-1980's. From the beginning, there was a constant amount of bandwidth available, but there were no contents and no applications running on these networks. So we were commissioned to proceed with broadband application research for them. The first projects we did were mainly based on remote computing, because computing power was extremely expensive. At that time, we were working on the basis of two remote computing approaches. One was to send large databases over broad band lines to remote rendering machines and then wait for the results. The second was to send control signals from input devices to remote machines with a database to render something already stored and receive the rendered results back in real time.

One example of the first approach is a remote visualisation system for use in medicine. Doctors use it to send large amounts of computer tomographic data from a clinic over broad band to a remote computing centre, and then receive the three-volume rendered results.

So it was not real time at all, because the data were too large and were rendered in a batch mode.

An example of the second approach, real-time telecommunication, was the area shot of Berlin located at a teleconference in Tokyo in 1993. Users there could move their spatial sensor over the area shot. The sensor's position was sent over a telephone line to a real-time rendering machine in Berlin. And the rendered view from the sensor's position was sent back to Tokyo over a satellite link. So the user was flying around virtual Berlin and Tokyo, but the pictures were rendered in real time in real Berlin. We used this system to visualise the different planning proposals we made at that time for the new centre of Berlin and presented them with this technique on several occasions.

* * *
Terravision's Rendering of the Earth
We would not use either this remote, real-time approach or the previously shown better approach today, because the cost of computing power and bandwidth are continuously dropping and at the same time, the worth of information is growing. It was due to this that we shifted from remote computing to the organisation and integration of decentralised network information. Our first project in this field was Terravision, which we started two years ago. Once we had decided to create this distributed earth visualisation system, the first technical task was to develop a renderer capable of visualising a database with unlimited geometry and texture in real time. The only thing we knew for sure at that time was that the earth is bigger than your main memory, bigger than your file system, bigger than a 32 bit address space and bigger than the precision of a 32 bit float. To sort this out, programmers have developed a renderer of which asynchronically loads geometry and texture level of detail out of databases distributed around the world.

This sounds complicated, but their is a very clear concept behind it. When the user is navigating around the virtual world, the renderer receives all flight parameters, such as precision and direction from the input device. From these, it calculates the currently needed data and predicts the needed data for the future flight part. Than it acquires them at an appropriate resolution from the responsible database somewhere in the world and loads them over high speed connections into the memory.

The closer we come, the more detailed the data becomes. Any data outside the current viewer is clipped away. If the viewing distance to the surface increases, higher resolution data are removed from memory. We thus always have the same amount of data in memory and are able to come infinitely close to the surface of the earth.

The database is accessed by the renderer asynchronously without affecting the frame rate. For instance, if the user approaches too fast or if the connection to the surface is too slow, you will receive a coarse image but the frame rate is constantly 30 Hertz. The renderer doesn't stop to wait for the necessary high resolution data and instead continues to fly past using the low resolution data until the other data required have arrived.

For the rendering, we are using a SGI Onyx and for networking, we are using ATM broad band connections. The T-vision texture data base is generated from various data sources, such as satellite imagery and area shots. We use source maps for calculating small and fast texture patches. To get smoother dissolves, we are also generating intermediate levels between the different levels.

The topography database, like any other database which could be integrated in the system, is pre-computed and organised using the same principle. This example here shows the texture database that we as a local Berlin provider have pre-generated and which we made accessible to all T-vision users in the network.

We hope that in the future, many pre-generated databases will be locally provided by responsible institutions or private individuals. Because this is not organised yet, except for demonstration purposes, we built a locally running version.

* * *
Mirror
Advancing technology has provided mankind with constantly updated images of the earth as seen from space. These images are a mirror of our responsibility for the range of ecological disasters caused by humans. The depletion of the ozone layer, including the advance of deserts, the destruction of the rain forest and damage caused by acid rain. Beside these images, there are many sources around the world which offer supplementary data like statistics or value-added data such as forecasts. It is neither sensible, nor possible to store all this in one place and most of these data can be read only in a responsibly visualised form. For this reason, we have developed T-vision as a tool and medium for generating knowledge from information provided by local services across the world. This is knowledge about the condition and the possible future development of our planet. In the best case, it will also influence our everyday behaviour.