I tried to do something very similar (except all the weird shooting your face and ambient lighting stuff of course) but the system has flaws that can't be overcome easily. Also there wasn't anyhting like kinect yet. It'd have some basics easier.

First of all I found projectors to be useless.
One of the biggest problems with projectors is that it's almost impossible to "feel" immerged. Yes, people believe otherwise because this is excitingly new to them and because they like the size - but after a while you notice the flaws.
With projectors you will always just stand in front of a wall. Imagine you go up close to a car in game and look though the side window, actually having to bow down a bit and aim lower, but all you really see is a GIANT window displayed across the room a few meters away from you and your brain says: what the hell? Not immervise.
Same goes for all situations of cover. You hide behind rock but there is no rock in front of your feet. There is that image 2 meters away from you. You brain again laughs at you hard.
Another key word here is depth of field although, interfaced corretly, this could be overcome from within the game.


Secondly there is the issue of aim tracking.
Aim tracking is bad. ArmA does much better in that regard. Humans do not rotate to look somewhere. We do 90% of the looking with our eyes and head movemennts. Even during body movements.
This is more or less an issue of the game though.


This said I put my project on hold for the following reasons.
- The inability to simulate anything beyond walking of flat concrete.
No matter how good you think it is. It's not. It's just outright weird to climb stairs, drive a vehicle, swim, ride, climb and even jump. Humans use all sorts of natural helpers to achieve their goals. For example we usually don't JUMP over a rock - we climb up and down. Even something as small as a fallen tree. We often put one foot on top etc. Leaning against a wall and looking around the corner? Good luck with that.
This will be extremely tough to overcome and this made me lose interest. It's literally that after a while you'll get annoyed by your body rampaging against all this abuse and your senses getting distorted.

- Better HUDs (head up displays) are neccessary in my opinion. Higher resolutions and eventually the ability to: track eye movements and most importantly: focus.
I could be motivated to start with this again as soon as they hit HD resolution without eye and focus tracking.

- Better tracking
tracking is in a sad state as we all know with natural point literally holding back all progress. Kinect may solve some of it but we still need high precision head tracking - something track natural point doesn't even intend to fix.
A real tracking input solution is needed. Flexible and generally accepted. NP with their weird infrared crap are totally holding everyone back. What about video? What about gyro? What about laser? There would be many other methods of tracking that would allow FULL 360, no, 3D tracking. Even FreeTrack could be changed to do this but we all know how that went.

- Better positional audio that hooks into tracking
Like someone said above you need to find a way to make fine adjustments to the sound axis. This could eventually be a software solution or an external hardware mixer. A software solution that hooks into the tracker would have the advantage though as it'd be less complicated hardware and thus could rely on one powerful and precise tracker that even knows the angles of the head itself (pitch and roll most importantly, yaw being obvious) as even the angle of the head changes the hearing, not just the rotation alone.



With kinect we have some powerful tool on the horizon. Sadly it's proprietary as TrackIR although Microsoft is very interested in pushing it forward while NaturalPoint has forgotten about innovation I guess.I assume it will greatly help resolving MANY of the issues I had.
For example controlling the treadmill speed and such. Imagine jumping forward with a leap. In real life you just touch the ground. In the sim you are greeted by the treadmill running at hefty speed to bring you back inward. Not only will you fall over but also you get the feeling of a car braking as backward inertia is generated.
I suggest 3meters as the radius for the mill and dynamic motion detection e.g. by the kinect. Jumps thus need to let you land and the slowly pull you back for example.



All in all it's becoming more reasonable to try this but it will still fail either due to the games, hardware or patents holding the industry back. In 5-10 years, after the 3d hype, I believe we'll see much more of this as holographic interaction is a focus of Mircosoft and thus kinect will certainly evolve into a very precise recognition device.
Considering the investments I'm still not motivated enough to continue my work on this. Seeing how much they want to sell this for is almost funny considering how disturbing these flaws felt to me.

Hmm, what about something like Sony's new HMZ-T1 with head tracking built in (freetrack could work easily, just mount a few IR leds), and somehow adapting the principle behind PS3's Move device for tracking a gun point? That way you can have a fixed camera and variable gun point.

Uh guys, your thinking about the movement platform backwards. That is not a motor. There is no motor. This is an input device, not output. i.e. the gears turn when you walk. Something reads the rpm of the gears as you walk causing the belts to turn.

I assume the IR detects what piece of the "pie" your in and then reads the rpm of the system and uses those two pieces of information to calculate direction and speed of movement. Their engineering on using a single rpm for the entire circle is nice.

Their price is probably a little high for their target market (military sims) but WAY high for commercial purposes. The reality is a bunch of inferred cameras, some slightly customized cargo systems (rollers, belts etc) and some custom software.

You could reproduce this with some parts from Conveyor Systems

Like all things simulator, ita the software integration that's always the hard part now a days.

In other words, although all those rollers look cool, the IR cameras do 80% of the work. The fancy platform is just there so you do not walk out of range of the cameras!

Nope, you would need a slope that increases toward the edge, and still the friction would be too great (through gearing and linkage) to "auto center" the player.
The player will be able to walk off the platform and the blue unit is too big for a sensor and is the size of an electric motor.

Incorrect. The blue device is a motor, and it turns all of the conveyor belts under the rollers at the same speed simultaneously. The IR cameras detect how far the user is out from the center, and adjusts the speed of the motor and thus the conveyor belts and rollers based on that and probably the speed the user is moving forward at. This is to maintain a position relative to the screen and re-center the user. With the rollers on flat ground and level, like they clearly are, even if they had amazing bearings, it wouldn't take much to walk off the edge if the rollers weren't pushing you back.

Looking through the pictures at how the whole platform is constructed, there can be no other explanation. The rollers sit on top of a conveyor belt, which is connected to the conveyor belts to it's left and right via a rotating shaft, and eventually to the blue box in question. The amount of friction that all these interconnected pieces would create means that by stepping off the center pedestal without the device operating, you would have no problem walking to and off the edge.

Basically, what JAMF said.

I'm guessing that the IR cameras do double duty in that in addition to controlling the speed of the rollers based on where the user is, they also provide the directional input based on the same data.

What do you think that the infrared-camera used? and what power the engine?