I don't know much about this space, how do you know it's definitely not positionally tracked? My reason is that they didn't talk about it, so I figure it isn't but I'm not certain.
None of the controlled objects in VR responded to any of the positional movement. You can see that the wand / pole is just rotating around a center-point attached to the user.
Positional tracking is a really challenging problem with a lot of limitations. If they were using computer vision to track the controller it would need to be held within the device's FOV and have identifiable elements such as a marker or LEDs. There are other approaches, but nothing I've seen that would work well on an HMD.
I've been working with AR/VR for the better part of a decade, and controller tracking's come up several times. I can tell from my own work that the device they're showing off is IMU only. It's too bad, something equivalent to the Vive controllers for mobile VR would be fantastic.
And at the end of the day - you're right. If it was spatially tracked then they definitely would have mentioned it.
> Positional tracking is a really challenging problem with a lot of limitations. If they were using computer vision to track the controller it would need to be held within the device's FOV ...
First, I think positional tracking of the person within his/her environment is one of the main things that makes the difference between VR and AR. So there's value in that. If you can do it with computer vision maybe it can work in almost any environment.
Second, there are RF based positional tracking systems. They're more expensive, true (not if you can design chips yourself, so might even be cheaper for google), but you don't need the controller to have visible elements.
Third, hand tracking like this should be able to work usefully, shouldn't it ? I know it doesn't do distance, but ... Just hang it off the viewer pointing downwards with a large volume immediately in front of the user. Not great for gaming probably, but for a virtual keyboard it should work, no ? http://www.ctxtechnologies.com/products/vk-200-keyfob-virtua...
Because good enough IMUs and optical systems for positional tracking are hella expensive compared to where in the market this thing is targeted.
There'd need to be either:
- a very expensive IMU in the controller, talking $300-500
- a camera and hefty CPU in the controller,
- a camera and dedicated image processing CPU on the headset
Vive positional tracking works because you control the environment. Oculus positional tracking works because you're already running on a hefty PC. You don't have either of those advantages on mobile, so you're left with either IMU sensor integration or SLAM, both of which are more expensive than $100 to do well and leave processing time for graphics.
The Gear VR/Oculus IMU is only good for orientation. Positional tracking with the Oculus Rift is done with an external camera, detecting a constellation of infrared LEDs on the headset itself. Gear VR thus does not have positional tracking.
To be able to integrate the acceleration data from an IMU to get positional data out of it is not something you can do with commodity IMUs, even the one made for Oculus. They're too imprecise and too inaccurate and too slow, and any analysis tricks to improve the results all introduce latency. Oculus' sensor has very little drift (though it still has drift, which is another thing the camera system on the desktop Rift can correct for) and the noise in the orientation data is not noticeable or can be filtered much more simply without introducing "too much" latency.
There are IMUs that are good enough to integrate acceleration data to position data, but they are prohibitively expensive for this application.
For what it's worth, I used to see a marked difference in the quality of orientation tracking between my Galaxy Note 4 in Google Cardboard vs. the Gear VR, but I see no different between GC and GVR on my Galaxy S7.
