Positional tracking already!

Looks good! But without any details about how he did it and how well it works, this is just a kid screaming into the camera.

For me it's more about the very small movements while sitting down and not walking around in a room. I think the video shows very well why this is not very practical at the moment. I'd be interested to know how these slight movements translate into the demo with the Hydra.

If this is actually a viable solution I will go and buy a Hydra right away.

Even if you don't move your entire body it is an improvement.

Really would depend on how accurate the hydra is in practice.

The marketing materials imply an accuracy of 1 mm for position tracking, and 1 degree for rotation.

Taking that at face value, 1 mm is pretty accurate, but conversely, 1 degree of rotational accuracy isn't.
- Although it would seem to benefit from giving you an absolute angle rather than a rate of rotation, the lack of drift aside, that's pretty bad.

By comparison the Wii motionplus gyroscope operates in a slow or fast mode, with 14 bit accuracy. Slow mode is equivalent to about 400 degrees/second for the maximum reading...
Since it's a signed value, that's about very roughly 8000 values between 0 and 400 degrees/sec or about 1/20 degrees/sec.

Of course, that's all a bit of a tangent, because the occulus already has rotational tracking, so it's really the accuracy of the linear tracking that's at issue.

Clearly, there's some room to question whether the hydra can actually report it's position as accurately as it claims, but if so...

Well, 1 mm is a pretty small distance. That's about on par with the kind of distance my head moves just due to breathing in and out.

Still, while I have a hydra on the way, I'm more inclined to stick a set of IR leds on the front and use the infrared tracking camera in a Wii for position readings as a first attempt.

I have no idea how accurate that is, but I do know you can get full 6 dof tracking from the IR system alone as long as there are 4 IR leds visible. - Latency then becomes another potential issue though; - it tracks at 100 fps, but to extract 3d coordinates and rotation from it takes a bit of processing work...

Wait for new hydra with 3 sensors (one for headstrap)! It will probably be announced VERY shortly!

Let's hope oculus and razer can work together here. It will be hard. I can forsee razer releasing theirs, but if oculus does it's own, it's gonna mean bloody mess for developers. We need a standard!

Heh, well conceptually, the data you're getting is pretty standard.

Orientation and position.

As long as you can design around some minimum standards of accuracy, you should be able to abstract away the hardware involved reasonably well.

But those minimum accuracy standards are the real problem...
Conceptually I could see using the Razer Hydra, Wii Remotes or Playstation Move controllers interchangably.

Unfortunately, the reality is the tracking accuracy and individual quirks of those 3 are quite different.
About the only thing you could really standardise amongst them with any degree of reliability is 3 axis rotation tracking.

The Wii can theoretically give position tracking using a sensor bar. (a non-standard design would work better than Nintendo's system; at least, for 3d position tracking), but the field of view in which tracking works is very limited.)
Alternately, using a remote as though it were a fixed camera, you can place a marker on an object (such as the rift headset) and track that in space.

Move tracks 2d position relatively well, but is much less accurate for depth sensing. It also has issues with occlusion of the tracking orbs.

The Hydra is, from what I've heard, prone to some degree of magnetic interference, and issues with tracking at close range to the base station. (Also the wiring could cause issues.)

Still... Abstraction would go a long way... IF you could deal with the variations in accuracy.

Carmack spoke about this in an interview at E3 2012

The issue is apparently that the Hydra is precise to 1mm but not accurate to any useful degree.

You'll move your head and end up with your view at a funny angle or you'll move slightly in a different axis.

Although that isn't an issue when you're using it in your hands like a controller but when controlling your head position I can see how it would be disorienting.

Hmm. It sounds like it's got some really odd quirks.

But it might depend on the actual nature of what goes wrong with it.

If the positional tracking is independent of the rotational tracking, then it's a fairly trivial situation.

If we can rely on a movement of 70 mm to the left, and 90 mm up being accurate, but the orientation being off by a considerable amount, then we can simply over-ride the orientation information using the tracking built into the rift.
After all, rotation data is what we already have. Position data is what's missing.

If however, position and rotation data are inter-related somehow, (a situation I can certainly imagine being the case), then the error in orientation tracking would also throw of the position tracking.
If that were the case, the hydra becomes almost useless for position tracking of head motion.

Still, like I said, even though I have a hydra on the way, my first inclination is to use IR tracking with a wii remote to handle head movement, and leave the hydra entirely for hand movements.
(I might even drag out the balance board while I'm at it. XD.)

Regardless, I'll be testing the tracking issues of the hydra as soon as I have one, so I'm sure I'll find out which situation we're faced with.

edit: I love how in this interview Carmack claims the rift kits would be in the region of $500-600, and that's for unassembled versions... And he thought that was cheap... Considering the final price is about half that... I really don't know what to say. XD

"IVI4tt" wrote:
Carmack spoke about this in an interview at E3 2012

The issue is apparently that the Hydra is precise to 1mm but not accurate to any useful degree.

You'll move your head and end up with your view at a funny angle or you'll move slightly in a different axis.

Although that isn't an issue when you're using it in your hands like a controller but when controlling your head position I can see how it would be disorienting.

He was wrong because the sensors complement eachother: accelerometer and gyro gives good angles (otherwise the rift wouldn't work at all) with magnetometer to correct yaw drift.

As the hydra has 3-axis magneto-thingy with base station (the ball dock that rests on the table) which gives position:



I suspect this might be used to further reduce drift in pitch, roll and yaw.

It's not perfect, but it's the best we got so far.

Edit: seems the solution is to put the base station om the HMD:



And have a third sensor on the table?

But since this photo was in a post from August last year I'm kinda wondering; at least they obviously have been thinking about this for a while...

In my experience, Carmack was correct. At least with positional tracking, the Hydra is not accurate in its readings, but it is consistent. This means that if I place a Hydra controller into the exact same location relative to the base station I will get the same result (the consistency) but if I compare two positions relative to each other they will not match their real world relative positions (low accuracy).

When using the controllers in your hands to do stuff on screen this usually isn't an issue. You automatically adjust your hands so that your in-game hands are at the correct location. And I love the Hydra when used this way! :D Unfortunately, you would not be able to perform this automatic body correction if you're trying to use it for absolute head position tracking.

If you're just trying to move around and play with positional tracking then it may be good enough. In doing quick, gross movements you may not really notice that your head is not exactly where you think it should be. But if you're trying to accurately model someone leaning forwards a couple of centimetres to help alleviate motion/VR sickness then I don't think it is quite there yet.

@rupy:
In that photo they are actually using the PS Move controller on the head for their position tracking within the room. The Hydra controller is used for their hands.

- Dave

Yes, but why is the base station on the head? Is the precision better the closer to the base station the controller is?

If yes, they will probably have one base station for the head on the table and then another for the hands on the head.

I don't think a camera + UV-light solution is good enough; you need a straight line of sight between the two at all times.

Magnetic is the way to go on this. I'm sure they will improve the precision in hardware and software, guess that's the delay.