Conclusions

The first section will give an overview of the modifications made to the SVE library that were necessary to build the environments. In chapter 5, a list of requirements was given. The next section will look at this list again and see what requirements where met and what went wrong with the others.

The next two sections will answer the two research questions asked in the beginning of this thesis. The first of the two sections will look at whether it was possible to treat people suffering from a phobia by using VR. The second of the two sections will look at what was learned about presence.

The final section will give some pointers to future research and what is still missing.

SVE library

All three environments were built using the SVE library. A lot of my time during the summer was spent enhancing the SVE library. Now the SVE library is used in many different projects and at many different places. Some of these projects even use SVE without having anything to do with VR. This section will explain some of the major changes to the SVE library.

The SVE library was developed on an SGI Indigo Elan. This machine has hardly any capabilities for texture mapping. All its texture mapping capabilities are in software and this is to slow to be used in real time. The experiment used an SGI Reality Engine that has texture mapping capabilities in hardware. One of the first changes, and most noticeable to people on demo days, was the addition of texture mapping to the SVE library.

The models used in the environments were created on the SGI's using a CAD package called WaveFront. The next addition to the SVE library was the capability to read WaveFront models directly. A conversion was no longer needed from WaveFront to SVE. Now WaveFront could also be used to create objects with texture mapping information. Before, this information had to be added by hand. Also the shading information had to be added by hand. Now WaveFront is able to supply both the texture mapping information and the shading information.

The capability of reading the WaveFront models led to the next improvement of SVE. Now it was possible to read the color specifications of WaveFront, giving the possibility of having reflective objects of any color. After this addition, the objects that were created in WaveFront looked exactly the same in SVE.

A final necessary change to SVE was to add sound. This required a complete rewrite of the interaction between two machines and the sending of information across the network. This eventually led to the possibility of having one machine read the trackers, another read the keyboard and mouse events, a third computer for audio and yet another to create the image.

All this is documented in a comprehensive manual which enables novice users to use the library with just a little help in the beginning. All the changes I made are also described in this manual.

Requirements

This section will look in detail at the elevator and at the bridges and balconies, where these are different from the elevator. The requirements, as mentioned in 5.5, will be summed up again and with each requirement a small explanation will be given how we achieved it or what went wrong.

Software should run on a Silicon Graphics computer
Software should be built using the SVE library

These requirements were easily met since the only machine the SVE library ran on was an SGI computer. Other reasons for using the SGI was the fact that all the VR equipment was attached to the SGI's. Also the SGI's were the only machines that had enough graphics power to make the environments look real.

The prototype should be usable as a demo during the summer (June 1993 - September 1993)

The idea of creating the first virtual environment and the list of requirements was already made in the summer. Because of the late decisions, a quick first prototype of the elevator was built. This was visible in the extreme simpleness of the model. During the rest of the summer the elevator was improved, both the software and the models, and could be used as a demo during the rest of the summer. An exact number of people riding the elevator during the summer is impossible to give. A guess would be several hundreds of people. After the summer, the development of the other two environments was started. These were used later during demo days, but none of them has been used on a scale like the elevator. In fact, this is still the environment that is most often used to demonstrate to people the capabilities of VR.

The learning curve of the virtual environments should be small

The subjects should use the environments without spending too much time in them learning how to interact. During the demo days, all kinds of people spent time in the elevator. Most of them could interact with the elevator within a couple of minutes. Some people, however, never got used to interacting with the elevator. The elevator required the most interaction by the user. In the other two environments, the user had just to look around. Since most people could interact with the elevator, and thus with the rest, this requirement was considered to be met.

The experiments start October 1st, so the environments should be ready by then

The experiments were started in October, so this requirement was met. Although the last environment was finished the day before the experiment and was not tested.

At least two or three environments should be ready before the experiment

When we started to develop the software and the models, we hoped to have four or five models finished before the experiments started. The creation of the models used in the environments took up more time than expected. Finishing the elevator also took more time than expected. When the experiments started, we had three environments finished.

Audio should be used to enhance realism

Auditory input is often forgotten in VR. Correct use of audio can enhance the environment immensely. When building the elevator, we wanted to include different sounds, the sound of an elevator, elevator music. Even sound from the restaurant at the bottom of the scene. The problem was at that point audio was still missing from the SVE library. Some sound capabilities were already there and others were added. In the end, sound was only used to give feedback when the subject touched a button. When we started to at look how to set up the experiment, we noticed that the machine we were going to use had no sound capabilities. To still be able to use sound, we had to use another machine just for sound. At this point we discarded, with pain, the idea of using sound.

Applications should run at least at ten frames per second

To have an application that reacts instantaneously to the user, at least ten frames per second should be achieved. Because of this requirement, there had to be some consideration about what was put in the environment. The bridges and the balconies both made this requirement and ran at about 11 frames per second. The elevator, however, only ran at eight frames per second. During the testing, we found out that most people did not notice this slowness. They were in complete control of the environment. A lot of their attention was on looking around instead of noticing the slowness. When running the experiment or when we had people in the elevator environment for extended periods of time, they started to notice the slowness of the environment.

Applications should be written for a HMD with limited resolution

During the development of the environments and the experiment, the same HMD was used. During the development, we noticed that the best thing to do was use big objects, and make clear to people what they are going to see. The subject had an easier time finding the painting once told there was a painting in the hotel.

Treatment

In the first chapter, the following research question was asked:
Examine the efficacy of a treatment for a phobia by using Virtual Reality.

This question was only answered for acrophobia. Two experiments have been done to test whether it was possible to treat people from acrophobia. The results of these experiments were much better than what we expected. This led us to conclude that in the case of acrophobia, it was possible to treat people by using VR. A lot of other phobias also use Graded Exposure to treat people. We think that not only people suffering from acrophobia can be treated but all phobias which are treated by using Graded Exposure. Eventually we hope to be able to treat all people suffering from specific phobias by using VR.

With VR Graded Exposure, we also have shown that VR can also be used in psychology and even medical situations. Both these fields use highly paid professionals. Using VR, we hope to have found a way to save them time, and thus save the patient money. Also, VR can easily be used to train these professionals without using real patients. A final advantage to the patient is the possibility of having a special program put together for the patient using VR.

Presence

In the first chapter, the following research question was asked:
Identify and explore issues related to the VR concept usually referred to as presence.

Looking at previous work done about presence, this question was divided into five subquestions:

Is there a definition of presence that is sufficiently operational and quantitative to be useful?
What are the factors that create a sense of presence?
Are there subjective and objective measures that can quantify presence?
Are there applications for which a sense of presence actually improves operator performance?
Are there applications for which presence is a necessary ingredient? If so, how are these applications different from applications for which a more traditional display system is just as effective?

An experiment was proposed to treat people suffering from acrophobia by using VR Graded Exposure. This treatment led to five assertions:

Assertion 1: A person's experience of a situation in a virtual environment may evoke the same reactions and emotions as the experience of a similar real-world situation. This may be true even when the virtual environment does not accurately or completely represent the real-world situation.
Assertion 2: Each person brings their own Gestalt into a VR experience.
Assertion 3: A person's perceptions of real-world situations and behavior in the real-world may be modified based on his experience within a virtual world.
Assertion 4: A primary difference between the experience of an event in a virtual environment and the experience of the same event in a real environment is in the intensity or vigor of the experience.
Assertion 5: Familiarity with a virtual environment does not necessarily increase the participants sense of presence.

The treatment did work for acrophobia. This was only possible if we gave the subjects the feeling of height. This led to the conclusion that they had to feel present in the virtual environment. To be able to make a change as claimed in assertion 3, it was necessary to make the subject feel as if what he saw was real. This is possible if there is a certain level of presence. How to measure that this level has been reached, we do not know yet. The reactions of the subject are the same in VR as in the real world, as claimed in assertion 1. This led us to conclude that the level of presence is enough for the subject to make him feel as if what he saw was real.

An answer to subquestion 1 is a good and usable definition of presence. It is not possible to claim that the following is a good definition of presence yet, but it is one that can be used to get a better idea about what presence is. Notice that this definition is tailored to VR and does not contain any quantitative measurements of presence.
Presence is the sense of being physically present in a computer generated or remote environment.

With the treatment of phobias, we hope to have found a whole class of applications where presence is a necessary ingredient, subquestion 5. To make sure that it can only be done by using VR, more experiments are necessary. The same experiment could be done now but, instead of using VR Graded Exposure and no treatment, the WL group could be treated by using images on a regular computer screen.

This experiment helped us to get a better view of the enormous task that lies ahead of us. The five subquestions, together with the five assertions, can help us to create better experiments, that are better aimed at a certain aspect of presence. But it can be safely stated that it will be a while before we are able to measure the level of presence generated by a VR application.

Future research

To find out more about presence, more research has to be done. This research showed five assertions which need to be properly tested. Also, the assertion that it is possible to treat more phobias by using VR Graded Exposure needs to be validated.

To see if other phobias also can be treated by using VR, this summer (1994) an experiment was started at the GVU to treat people suffering from the fear of flying. During the summer, a prototype will be built and tested. After the summer, an experiment will be started where the prototype will be used to treat people. Again, the problem is that there is no knowledge of what causes the fear of flying.

Another experiment that needs to be done is to test how good VR Graded Exposure is, compared to in vivo and imaginal. I have the impression that it is slightly less useful than in vivo, but much better than imaginal. Again this is my feeling, to be sure an experiment needs to be done to test this.

Small experiments can be set up to test the assertions made. Important here is that, unlike the experiments described in this report, the assertions should be validated. Thus, an experiment has to be set up to test the assertions, not to see if the assertions come out of the experiment.

An experiment can be set up to test whether or not treatment of phobias is an application where the level of presence can only be reached by using VR. The experiment can be the same as described in this thesis, but instead of giving the WL group no treatment, the WL group can be treated by using images on a computer screen.

With this experiment, a small step has been taken to look at what presence is. Still, most of the subquestions asked in chapter 3 have not been answered. More research is needed to answer these questions. Answers to these questions are of course important to VR, but also to other fields. Movie makers for instance, want to create a sense of presence when you are watching a movie. User Interface builders want to create some sense of presence, so that the operator will choose the right option at the right time. This leads to the conclusion that much more research has to be done to find out more about presence, and that many different fields will benefit from this knowledge.

Rob Kooper

kooper@cc.gatech.edu

Last modified: Wed Aug 9 12:38:29 GMT 1995