Slashdot Mirror

← Back to Stories (view on slashdot.org)

Jack McCauley's Next Challenge: the Perfect Head-Tracker For VR (ieee.org)

Posted by timothy on from the it's-on-your-head dept.

Tekla Perry writes: He used a webcam and LEDs to do position tracking for the Oculus DK2, but Jack McCauley, co-founder of Oculus and now working independently, says that's the wrong approach. He likes the laser scanning system of the HTC Vive better, but says it's just not fast enough. McCauley thinks he can do better, using a design approach borrowed from picoprojectors. Speaking at this week's MEMS Executive Congress, he said better tracking of head position will solve the problem of VR sickness, not more expensive screen technologies.

25 comments

  1. I think everyone agrees by Comen · · Score: 1

    I think everyone wants better tracking, I mean I even think Oculus would agree that V2 might have something better, but this is a version 1, calm down things are all going to change and get better over and like comparing cell phone abilities. I am sure I will be getting a Oculus CV1 when it comes out, I really do not have the room for the VIVE myself, but I can not say I would not buy one in the future sometime on version 2 or something.

    1. Re: I think everyone agrees by Anonymous Coward · · Score: 0

      Vive can use the same area as the oculus or more.

      The oculus is inferior because it only allows one device in front of the camera, whereas you could use multiple vive headsets in the same room with only a single set of lasers

    2. Re:I think everyone agrees by Anonymous Coward · · Score: 0

      dk2 tracking works pretty well for me.

      he's right about higher refresh rates being totally unnecessary. if a few people want higher refresh rates then let them pay for it by selling them a special model.

      pixel density otoh does need to increase from the dk2 as although touted better than dk1, which I never used, I still noticed pixels almost immediately which greatly detracts from the experience, and also makes text and such difficult to properly render.

  2. "not more expensive screen technologies"? by jeffb+(2.718) · · Score: 2

    Sure, we need better tracking, with higher temporal and spatial resolution, and lower noise. But to get low-enough latency from position sensing to image display, you need high performance at every stage -- position acquisition, image computation, data transfer to your display, and display refresh rate. If the total of all those latencies exceeds the maximum tolerable delay, you lose. If you're going to get sick with total delays of 10 ms, and your display refreshes at less than 100 Hz, there's nothing you can do -- until you get a higher-refresh-rate display.

    1. Re:"not more expensive screen technologies"? by Pinky's+Brain · · Score: 1

      A 100 Hz display is probably good enough as long as it's stroboscopic and not sample and hold.

    2. Re:"not more expensive screen technologies"? by codeAlDente · · Score: 1, Interesting

      It's not just about latency. If your display is near zero latency and you don't account for head movements, you're going to cause sickness. That is an inevitable consequence of binocular vision. Latency is >10 ms in human photoreceptors, and adaptation to head position is based on vestibular feedback downstream. Latency is not the limitation. Matching self motion to visual motion is the limitation. If these are mismatched, latency won't matter and people will get sick. I applaud the dude for realizing this.

      --
      He once inserted random mutations into his code, just so he could have the experience of debugging.
    3. Re:"not more expensive screen technologies"? by Tyr07 · · Score: 1

      Actually human vision is very, very fast, and cannot be measured in terms of latency for organic visual acuity.

      Humans can see the difference between 1 MS response time and 10 MS response time.
      What the measurement is for human vision is not how fast it can detect a change, but how much has to change before we can detect it.
      E.G If you had one line of dots traveling at 1 MS and another line travelling at 10 MS - we would be able to tell there is a difference IF there was enough dots.

      Like half a screen moving to the right at 10 MS and the other half of the screen moving to the right at 1 MS.

      Analog vision is very different from electronic measurement.

    4. Re:"not more expensive screen technologies"? by Anonymous Coward · · Score: 2, Insightful

      What on earth do you mean by "traveling at 1 MS[sic]"? The units in question here are milliseconds (ms); they are not a velocity, or angular velocity, or anything that makes sense in the way you are attempting to use them.

    5. Re:"not more expensive screen technologies"? by KGIII · · Score: 1

      How would these work for not just virtual reality but for augmented reality? Is there a chance at putting a camera on the front of them and passing that through to the VR headset and then intersecting that with data? I'm thinking something more utilitarian... I'll try to think of an example...

      Say you're looking at a new house - you're there physically. You toss on a headset and scan the room by looking around. Then you put furniture in the room. Then you can move the furniture around, and check it out, and decide if you're easily able to make the room look like what you'd want. You can paint the walls, put in curtains, and look out the window. This needn't have the house's layout already programmed into it - just some stock furniture or other types of things.

      Something like Google Glasses but more useful and less spying. Err... Except they could be used for invasive shit too but, oh well... The above is just one type of example of things to do with it. I can think of a bunch more but the idea occurred to me while driving across the Skyway over to Niagara a while back. I was curious as to how old the highway was, how tall it was, what the buildings were off to the left, what ships those were when I saw a bunch anchored, etc... It'd be nice to be able to pull that information up - then mix it with VR, in some cases. So, more functional than just Google Glasses and more immersible.

      Also, no, obviously one shouldn't wear them while driving. That doesn't negate the idea and doesn't mean a passenger couldn't wear them. It'd need some small attached horsepower but we can do that these days. Coupled with speedy connectivity via cell, it might be interesting. I think the realty idea above is a good example, they can walk over and move the furniture, move about the room, etc... I'm kind of assuming they don't already have a camera mounted on the front of them. I played with a VR headset a long time ago and it was horrific - this probably would have been like ten years ago at a convention. I'm assuming that they've progressed a long ways since then.

      --
      "So long and thanks for all the fish."
    6. Re:"not more expensive screen technologies"? by jeffb+(2.718) · · Score: 1

      I meant "position acquisition" in the general sense, which includes head position. As you say, head position is critically important, and we currently don't have any way to manipulate the vestibular feedback system. (And if you ever do develop one, good luck convincing users to let you mess with their inner ear.)

      The problem is that increased latency anywhere in the pipe translates to positional inaccuracy during slew operations -- the faster the slew, the bigger the error, and that's what leads to VR sickness. (Well, it's one thing that leads to VR sickness. Deliberately displaying motion that doesn't match what the inner ear perceives is the big nauseator, but that's hardly the fault of the hardware.)

    7. Re:"not more expensive screen technologies"? by codeAlDente · · Score: 1

      this is wrong. google flicker fusion frequency.

      --
      He once inserted random mutations into his code, just so he could have the experience of debugging.
    8. Re:"not more expensive screen technologies"? by codeAlDente · · Score: 1

      Interesting, my take-away from the article was that the display isn't being updated with the right information, i.e., perhaps different parts of the visual field should be updated differently due to the feedback from the head motion monitor, and current software isn't yet doing the optimal job given the inputs it gets. That was my impression from wearing one. But yeah, manipulating the vestibular feedback system would be a cool addition to any helmet. I think tcms or eventually optogenetics is up to the job.

      --
      He once inserted random mutations into his code, just so he could have the experience of debugging.
  3. How is laser scanning not fast enough by Pinky's+Brain · · Score: 1

    Throwing a horizontal and vertical laser plane across the space at 1000 times per second is pretty trivial, 10000 per second too ... 0.1 ms not fast enough?

    1. Re:How is laser scanning not fast enough by codeAlDente · · Score: 1

      You want to shine lasers into people's eyes??

      --
      He once inserted random mutations into his code, just so he could have the experience of debugging.
    2. Re:How is laser scanning not fast enough by Anonymous Coward · · Score: 0

      It won't be a problem if you're wearing a VR headset.

    3. Re:How is laser scanning not fast enough by Anonymous Coward · · Score: 0

      But what if the headset draws images on your retinas with friggin' lasers??

      Ha! There, I run rings around you logically...

    4. Re:How is laser scanning not fast enough by Pinky's+Brain · · Score: 2

      Sure, a 1500nm range IR laser plane scanned at a couple 1000 RPM with good interlocks. Why not?

      That range gets absorbed inside the bulk of the eye by the way, which is why it's slightly misleadingly known as eye-safe. Longer wavelengths dump most of their energy on the cornea and shorter wavelengths dump most of their energy on the retina, so they have much lower damage thresholds.

    5. Re:How is laser scanning not fast enough by Pinky's+Brain · · Score: 1

      On reading the article he just wants to use MEMS because supposedly that's necessary, which is bullshit. The only problem with larger scanners is that they might be a little noisy, 1000 scans per second with a polygon or small flat mirror is going to be quiet enough for a prototype though.

      I'd worry more about the detectors, doing sub ns time to digital is not really hard, doing it cheaply for a lot of detector is going to take a bit of R&D though.

    6. Re:How is laser scanning not fast enough by Pinky's+Brain · · Score: 2

      Oops, I'm a couple orders of magnitude off with the timing, I guess even 10ns resolution would be more than enough.

    7. Re:How is laser scanning not fast enough by grumbel · · Score: 1

      I think it's not so much that it isn't fast enough, but that you end up with a big box with a big spinning mirror. What he wants to do is to take the same technology and replace the spinning mirror with a DLP-style micromirror, thus you could do the same thing much smaller, cheaper and with less power, which in turn would allow you to have more of those devices for better tracking.

    8. Re:How is laser scanning not fast enough by KGIII · · Score: 1

      They'll just include sharks with lasers to fight on your behalf - they'll deflect the lasers with their lasers and you'll not have to worry about the lasers hitting your eyes. SOLVED PROBLEM!

      --
      "So long and thanks for all the fish."
  4. Perfect tracking won't fix lack of body movement. by Maxo-Texas · · Score: 1

    Maybe it will work if you are sitting in a chair in real life and in the game- but if the screen shows running and the body isn't moving, some people are going to still get sick.

    --
    She was like chocolate when she drank... semi-sweet at first and then increasingly bitter.
  5. Cardboard by Anonymous Coward · · Score: 0

    Cardboard has really good tracking. I got one for my Note 4 and it is amazingly precise. The Note 4 even has sufficient pixel density to make a semi-decent VR display.

  6. current hear tracking works fine by Anonymous Coward · · Score: 0

    It's the damn judder that makes you sick.