Oculus Further Details Ideal Hardware and Sensor Configurations for Roomscale VR with Touch

The truth doesn’t stop being the truth just because you were away LOL.

Well not to answer for him (whoever he is I don’t know, he’s a blocked user for me), I have heard these arguments given against lighthouse with regards to “scaling”:

There is a conjecture that constellation would scale to more tracked objects because it uses a different marker and tracking system than lighthouse. I do not fully understand these arguments but I think they go something like this:

– You can more easily add LED emitters than you can laser sensing diodes because the LED emitters are basically “dumb” and require little electronics in the tracked device
– You can more easily add LED emitters than you can laser sensing diodes because the LED emitter doesn’t need to send anything back to the base unit other than dumb light (no need for wireless)

Consider that the Oculus equivalent of the new HTC tracking puck would just have LED emitters and a battery, with no need for wireless.

However, consider also that such an Oculus tracking puck would be MISSING some features from the Vive puck such as the ability to transmit other inputs from the tracked device like trigger pulls etc. And to get those, which you’d almost always want in any device more sophisticated than a baseball bat, you’d still need to add wireless capability to the Oculus tracking puck.

Another reason that is proffered for why Oculus tracking is more scalable is that you can theoretically track any number of devices in the same scene since they’re all just clusters of LED dots.

I do not understand how this is any different from Vive though, which would just be clusters of laser tracking diodes, except that the Oculus solution doesn’t require every tracked device to be communicating via wireless to the headset/base station.

Once again, this is really only a benefit for “dumb” tracked objects. However, there are definitely some pretty “dumb” tracked objects that are important – like feet, wrists, hips, etc. That being said, I would really be surprised if adding Lighthouse tracking for feet, wrists, hips, etc, was significantly more costly or complex than doing so for Constellation would be.

Another way that the Constellation solution could be argued to be more scalable would be in terms of tracking volume — theoretically, with a high enough camera resolution and bright enough LEDs, you could track arbitrarily far out in distance, wheras the Vive tracking solutions sweeps lasers that would (possibly) lose coherence much more quickly at larger distances and thus suffer from a smaller maximum possible tracking volume.

However, Oculus’ inability to track even 10 ft by 10 ft well is some pretty good evidence that the practical realization of constellation is not nearly as good as its theoretical possiblities, and given the technology involved, I personally doubt that it ever will be.

I also am very skeptical that a Constellation tracking system could easily track hundreds of LEDs that are grouped into different objects with different orientations all occluding each other at different times. It sounds like an image processing nightmare to me.

Oh and let me add one more thing: one problem with lighthouse is that the two base units have to time multiplex their laser sweeps — i.e. one must flash a LED, then sweep lasers, then the other must flash a LED, then sweep lasers … you can see that if you tried adding another lighthouse unit that is visible to the others, it could only work if now each uses 1/3 of the available time instead of 1/2. Add another and each is using 1/4. This means that increasing tracking volume by adding more lighthouse units implies reducing tracking frequency of the whole system.

One way around this would be to spin the lasers faster, but I don’t know if that then starts causing problems at the diodes which can no longer see the lasers going by that fast.

My point about scaling (and I am not the first to ever mention this) has to do with what you perceive as scaling by just adding units. A solution that simply requires adding Lighthouse units ad-infinitum is not a really good solution ultimately. The problem really as Sponge points out is that any mechanical solution requiring multiple units is inherently limited for very large spaces (and utterly incapable of integrating in objects for an AR/VR application).

High-speed cameras allow a much broader coverage of an area (the lasers used for Lighthouse are just not that strong in comparison, but very good for smaller spaces) and not using mechanical parts means they are more reliable over time as well. The LED approach currently used by Constellation will likely yield to a different design but still incorporating cameras. Possibly Lighthouse will evolve as well, and not so many units over a space will be required, but as it stands, the need for a mechanical rig to spin to produce light is great for the current application but begins to show problems the larger the number and area covered. Even if they did get rid of the mechanical aspects of Lighthouse, it still cannot provide the additional AR/VR advantages a camera does. I find it telling that no one else, and I mean no one is going with a Lighthouse style solution I can think of, instead everyone is going with camera-based… maybe there’s a good reason for that over the long haul?

Check this link out, which acknowledges strengths and weaknesses in terms of barcode reading… please note however it points out one thing that the imaging technologies and image processing is continuously improving – that is why AR/VR seems to be tied more to imagers than lasers down the line: https://corp.trackabout.com/blog/lasers-vs-imagers/

But that’s why you add base stations and split your tracking volume up onto segments, each covered by a different set of base stations.

I seriously doubt that the camera based solution is ever going to have the fidelity to track small flashing LEDs at long distances. Ever.

What? Why do you need 2 lasers per player? Do you realize that two lasers cover the entire tracked space no matter how many items are in that space? And with the single-motor version of the lighthouse base stations you could probably go to 4 lasers in the space, which would improve occlusion issues. That’s all you’d need no matter how many objects you have in the space.

Please verify that you understand the basic principle of lighthouse technology, that the lasers don’t do the tracking, the lasers just provide a reference signal that the tracking diodes on the tracked devices use to do tracking.

Theoretically you could go to more cameras and have even better occlusion avoidance, but that’s theory, I doubt it would work well in practice.

You know, I had a feeling that you were trying to talk about that instead of Lighthouse. The discussion was about lighthouse. We’re not talking about prototypes of some unproven technology. We’re talking about how Lighthouse works and its merits when compared to Constellation.

You might want to read up how Vive works. True the laser can track 1 object but this is not how Vive works. It’s sole purposes is to light up the sensor in the HMD and whatever sensor the the object that can sense the light. So, this effectively voided your assumption that light house can only track 1 object. In fact, there have been demonstrated that multiple Vive HMD works with only 2 light houses which is impossible for Rift to work.

I know in long run camera has an potential but what we are discussing is now not future. With this kind of low bandwidth port and tremendously huge cost far high speed camera it is not practical at all now. If you want to discuss please discuss in context which is relevant, not fantasy. What if I tell you there will be a laser fast enough to scanner the object in the blink of the eye in future? Can it be used for tracking? Definitely, but how practical is it?
http://www.robots.ox.ac.uk/~mobile/Theses/WangThesis.pdf

I wasn’t talking about Vive

I was referring to http://www.roadtovr.com/mts-virtual-reality-vr-tracking-system-jack-mccauley-oculus-vp-engineering/

That thing can track 100 meters away (don’t look inside laser though) BUT only ONE object

They forget how much bandwidth is required to push thru. the USB cable the need to tether back to PC.LOL.
They start to fantasy and imagine thing that far beyond the current gen. VR and USB limitation.
Just the Rift tracking up now and stop bullshitting on how good the camera can be in future.

Very good point….

Can the laser bring in the surrounding objects? No, didn’t think so and that is the point… its about more than tracking…. why this unfounded bias towards Lighthouse? It’s great but it has limits, and potentially one fatal one in time… why this is an issue I cant grasp… I plainly point out the flaws in Constellation today, why can’t people accept the limits of Lighthouse? it’s not perfect, nor the solution for every case….. no solution is, nor is any solution perfect…

As far as ensuring IDs are unique, they could be configured in a dumb device by an external switch or switches easily enough. Barely more component cost than the board that pulls them in the first place. A set of four dip switches next to the battery gives you sixteen IDs, which, I believe, doubles the 8 addresses (so 7 devices plus the headset) that Vives Bluetooth PAN protocol currently maxes out at.

As Bryan points out, if you start tracking feet, hips, and so on to get skeletal references, you run out of addresses on Vive quickly.

Consider also that a future camera based tracking setup would infer the room layout for you, and that VR will be able to borrow heavily from research and advances being made for AR systems on the inside-out camera-based tracking front, and I see camera tracking outliving lighthouses in the long term.

Thanks for the correction on the addressing space. I thought it was 3-bit, not 4. I also agree on the broader point you make about tracking inert devices.

I think that 2016 is likely to drive some interesting trends in the mid-range future (18 months to 3 years from now, probably) as far as hardware development goes. The bandwidth issues Oculus is running into are primarily motherboard and chipset manufacturing habits. USB 3 controllers are expensive and not in particularly high demand for the consumer market — previously the only real benefit over USB 2 has been external hard drives, and most users only need one of those.

Kaby Lake and the 270 line of chipsets just came out, and were thus too far along in the pipeline to react to this, but I feel like the next major chipset revision we see from Intel is likely to make a concerted effort to step up the offerings on PCIe lanes and USB 3.1 support.

static hi-res camera is better than space scanning mechanically rotated laser beam. period.

That’s the problem – you only understand the way it all works by comparing laser vs camera…. as an entire system it is more than that, simple minds fail to grasp the long term direction of AR/VR.

Lighthouse is a *very* good system for small areas, and only if your goal is to track objects. Period. The camera solution, while on the surface creakier at first, offers a lot of advantages over time such as bring the user into the space, or other objects. The Lighthouse method cannot do this, and requires tacking Trackers on anything but then generating the associated image inside the space by hand which works fine for gaming in VR with limited objects, but is far less flexible a system over time.

Cost is always headed down… thats actually one of the good things about it… as it improves technology costs also go down.. when I say good for small area let me clarify – a 30’x30 area is “small” if one compares it to what high-speed resolution cameras might cover…. and we are talking about near future developments not today. I am very consistent in my claim that Lighthouse is a superior solution today, and for a larger area than Constellation….