So I'm going to give a brief overview here and then we will go into each of these areas in more detail in the coming weeks
We talked about several display options last time, from Fish Tank VR, through Head Mounted Displays and BOOMS to projection based systems.
It is important to note here that although the field is called 'virtual reality' the goal is not always to recreate reality.
Computers are capable of creating very realitic images, but it takes a lot of time to do that. In VR we want at least 15 frames per second in stereo (w/ roughly 1 million pixels per monoscopic frame).
For comparison:
The tradeoff is image quality (especially in the areas of smoothness of polygons, antialiasing, lighting effects, transparency) vs speed.
Though in some cases, like General Motors, they sacrifice frame rate (frames per second) for visual quality.
If we want stereo visuals then we need a way to show a slightly different image to each eye simultaneously. The person's brain then fuses these two images into a stereo image.
One way is to isolate the users eyes (as in a HDM or BOOM) and feed a separate signal to each eye using 2 display devices. Each eye watches its own independent TV.
Another way is to use one display device and filter what each eye sees. There are several different ways to do this.
We can use colour - this has been done in 3D theatrical films since the 50s with red and blue (cyan) glasses.
We can use polarization (linear or circular) - linear polarization was used in 3D theatrical films in the 80s (and to be totally truthful, many of the better 3D films of the 50s were done with linear polarization first)
We can use time - this was common in VR in the 90s and continues today where we show the left eye image then the right eye image and the user wears LCD shutter glasses which ensure that only the correct eye sees the correct image by going opaque on the eye that should be seing nothing. These glasses used to cost over $1000 each in the early 90s. Now it looks like they will be the basis of the new 3D digital cinema coming in the next few years with supposedly costs of around $30 per pair.
In all these cases both of the eyes are focusing at a specific distance - wherever the screen is located. There is no way for the user to change focus and bring parts of the scene into focus and let others go out of focus as in the real world .
"people hate helmets, but people like sunglasses"
ergonomics and health issues of various displays
Typically museums and other places with many visitors it is necessary to either give the glasses away to the user (with the paper ones) or wash them (with the polarizing ones) to keep things sanitary. This is more difficult with HMDs where people have tried using alcohol wipes.
Need a computer capable of driving the display device at a fast enough rate to maintain the illusion.
In the past (ie the 90s) that usually means either simple scenes, very specialized graphics hardware, or a lot of work in optimizing the software. But this is less true today where scenes are getting more complex, the hardware more commonplace, and the software more capable.
Benchmarks on CPUs and graphics cards aren't really very meaningful. They can give ballpark figures but there are a lot of factors that combine to give the overall speed/quality of the virtual environment.
Multiple processors are usually required, since there tend to be multiple simultaneous jobs to be performed - ie generating the graphics, handling the audio, synchronizing with network events.
Multiple graphics engines are pretty much required if you have multiple display surfaces
Ability to 'pipeline' the graphics is pretty much required
At minimum you want to track the position (x,y,z) and orientation (roll,pitch,yaw) of the user's head - 6 degrees of freedom.
You often want to track more than that - 1 hand, other hand, legs?, full body?
An Important factor is how far the user can move - what size area must the tracker track?
Can line of sight be guaranteed between the tracker and the sensors?
What kinds of latencies are acceptable?
Input devices are perhaps the most interesting area in VR research. While the user can move their head 'naturally' to look around, how does the user navigate through the environment or interact with the things found there?
Ambient sounds are useful to increase the believability of a space
Sounds are useful as a feedback mechanism
Important in collaborative applications to relay voice between the various participants
Spatialized sound can be useful
Often useful to network a VR world to other computers.
We need high bandwitch networking for moving large amounts of data around, but even more important that that we need Quality of Service guarantees, especially in regards to latency and jitter.
most VR programming is done 'from scratch' or using lab-based software. There are a few major commercial products in use: dVISE, World Toolkit and several open source libraries.
CAVE
Varrier
PARIS
As with many scientific 'do all' solutions, VR is at the point where it can no longer be sustained by hype. VR is still immature, but very promising. Unlike many of the other 'do all' solutions that got hit with a serious backlash, VR has the advantage of being obviously 'fun'.
Tracking