Position tracking for a seated user - not that hard?

[rmcclelland]

Watching all the Oculus Rift CES demos, it occurred to me that position tracking (x,y,z) may not be that hard if the user is seated. Until the virtual locomotion problem is solved, I image most people will be seated for VR experiences. If you are just trying to capture head motions such as leaning in or looking around corners, which seems to be the immediate issue, the range and occlusion requirements are greatly reduced.

If you combine the orientation data from the Oculus Sensor (rx, ry, rz) with the position data from a PS Move/Eye (x,y,z) doesn't that give you everything required for 6dof head tracking?

Looking at the beginning, translation only, part of the video in @Fredz post:
http://www.mtbs3d.com/phpBB/viewtopic.p ... s+move+api
The position tracking seems great. The author of the API claims the end-to-end system latency is 68ms, which isn't great, but isn't terrible either. Perhaps sensor fusion with accelerometers would help.

The sphere would need to be placed somewhere it would not be wholly occluded. Perhaps right on the front of the visor? It wouldn't work if you turned completely around, but that is a limited use case. You can't easily swivel your head past 180 degrees around while leaning.

Thoughts? I guess I'll have to pick up a PS Move/Eye and wait for my Rift. Not sure how game integration would be handled until the SDK is out.

This also seems to be what they are doing with Project Holodeck:
http://www.projectholodeck.com/wp-conte ... acking.png

[cybereality]

Sure, it will work. As long as you don't mind having a light bulb taped to your head.

Could also probably get some IR safety goggles ala Johnny Lee and use a Wiimote for positional tracking. Would limit your sweet-spot more than the Move, but I imagine latency would be lower and there is probably more software floating around.

[rmcclelland]

Then I would always look like I had an idea...

Thanks for the tip on the wiimote. Lots of resources for sure.

It seems like it would be pretty easy to make a position tracking add-on for the Rift. I image a few of these will come out quickly.

[brantlew]

Yes, I would say the problem for unidirectional seated play is greatly reduced from the general problem. Even the Hydra is a solution then.

[MSat]

It's definitely easier when you place limits on what movements the user should be able to perform. Just as a note, handling the limited motions of someone standing in place is similar to a seated user, so any given system should work for both. With current technology, position tracking would utilize either an IMU, optical system, magnetic tracking, or a combination of any of those.

I think the simplest route is to use an IMU on the upper body - at least as long as the user pivots around a fixed spot such as the waist, but keeps their spine rigid except for twist. The problems with tracking while seated using this method would only arise if the user pivoted their upper body around their waist, while keeping their shoulders level (though moving the tracker to the mid-section might alleviate this). As for standing, a problem arises when a user keeps everything above the waist level, but pivots the lower section around their hips.

A possible solution for standing game play may be to sling a camera over a user's shoulders behind them pointing down - looking at a "play pad" with some sort of trackable pattern. If there's an accelerometer level with the camera to help determine if it's at an angle relative to gravity, translations and orientation should be resolvable - at least as long the user doesn't go out of bounds, or leans forward too much.

[rmcclelland]

Sure, it will work. As long as you don't mind having a light bulb taped to your head.

Could also probably get some IR safety goggles ala Johnny Lee and use a Wiimote for positional tracking. Would limit your sweet-spot more than the Move, but I imagine latency would be lower and there is probably more software floating around.

I don't believe the "Johnny Lee" method would work. As I understand, it uses a single Wiimote to look at two IR LEDs. x and y are the average x and y pixels of the two light points in the sensor plane. The z information comes from the separation between the two points since he is always looking directly at the Wiimote. For VR, when you turn your head left or right, the distance between the points would decrease as projected onto the Wiimote sensor plane, making it think you are moving your head away.

I think you would have to use two Wiimotes, which is a bit more complex, as shown below:
http://www.youtube.com/watch?v=CT6aQN-l ... PL&index=9

[morgg]

You can build your own 3 LED marker (similar to TrackIR's Track-clip), and use Freetrack for 6DOF tracking.

In theory, you would discard (pitch, yaw, roll) and just take (x,y,z) to combine with the orientation data coming from the Rift.

I don't know if there would be some complicated geometry considerations when combining the two independent sensors.

[rmcclelland]

You can build your own 3 LED marker (similar to TrackIR's Track-clip), and use Freetrack for 6DOF tracking.

In theory, you would discard (pitch, yaw, roll) and just take (x,y,z) to combine with the orientation data coming from the Rift.

I don't know if there would be some complicated geometry considerations when combining the two independent sensors.

Thanks for the tip on Freetrack. Sounds like the best solution. Also sounds like Palmer/Oculus are going this route:

"3 months is not nearly enough time to work out something as complex as proper positional tracking. Copying something like FreeTrack would be hard enough, but adding a prioritized optical/inertial sensor fusion system would have been impossible, especially being as small of a company as we are."

You should be able to treat the head as a rigid body, in which the 6DOFs are independent. You would have to replace the head/neck model with a model of the HMD/trackers.