6DOF settings for enhanced positional tracking

Pimax Play Discussion
1

I have already shared why I find CV1 IR camera tracking is superior to LH tracking.
I won’t come back on this here but want to address a specific area in which I was getting better tracking with my CV1 than with my Pimax.

Here is what happens when I leave my 5K+ laid somewhere, with 6DOF tracking:

Ideally there should be no shaking at all.

  • This happens with 2x v1 LH (seeing each other), 1x v1 LH, 2x v2 LH (no need v2 see each other), or 1x v2 LH.
  • LHs firmly attached to walls (also tried with LH just laying onto bookshelves)
  • HMD plugged to usb3 or usb2.
  • Each time the LH is/are visible from the HMD.
  • happens with both my first KS 5K+ and RMA replacement (sn203)
  • I was not able to improve this by making other faces of the HMD face the LH(s), nor by trying to remove any reflective material nearby, nor by trying to move the HMD to another place within the area covered by the LH(s).
  • LH is set at about 1.8m height, pointing down.
  • I have not tried optimizing LH placement for hours but I find tracking should be flawless with current LH placement.

I think the positionnal data should get some processing to eliminate this kind of shaking.
It may be eliminated completetly without any noticeable loss in precision nor latency.

I imagine a processing like this:

When HMD movement speed is very low (=almost steady head), do some temporal averaging and add a decreased precision on top if needed (set both so that a laid HMD provides a perfectly steady view).

Then, as HMD movement speed increases, progressively do less temporal averaging (to avoid any feel of latency). If needed I guess precision could be decreased more, until the tracking is felt smooth (=comfortable), because you don’t need super high precision on fast movements.

It should just be a matter of finding a good balancing for those parameters (temporal averaging, precsion decrease, and their respective scaling with movement speed). Pitool could offer sliders to let the user find the best settings for him.

I think this could noticeably improve the comfort from pimax VR experience.

And maybe this idea could also be applied to controllers tracking: have you never thought your VR controllers felt too shaky ? Especially if your hands are already shaky IRL. Maybe your hands could be made smooth in VR without feeling fake/unprecise ?

@SweViver

2

After writing this post I have googled that issue, found a vive user with exact same issue who solved it by plugging the HMD to usb2 instead of usb3. Doesn’t fix it for me with the 5K+.

Tomorrow will check if OG vive shakes the same or not.

1 Like
4

This is simply not true.

I have been using the CV1 with 3 cameras and the tracking was flawless, almost 100% perfect.

Too close ? This is LH tracking that doesn’t work well when being too close, at least the 5K+ isn’t even being detected until you pull it far enough from the LH. My 1st unit wouldn’t even detect until I show the side/rear of the HMD to the lighthouse (was also reported by other pimax users here). Seems a bit better with my sn203 replacement from RMA.

At the opposite the CV1 tracking allowed me to have the camera very close, I began with a single camera right on top of my monitor for racing games (was of course suffering occlusions with a single cam). The camera was about 1m away from the HMD.

If Pimax went for IR tracking you would be able to manufacture your own cameras today instead of having to rely on valve to source LHs.

Oculus didn’t drop IR tracking because it would be technically bad but because they went inside-out for their mass market strategy.

Have you watched the video above ? Do you find a shaky tracking like this is good ? :smile:
And if it was the only issue I already had with the 5K+ and its LH tracking…but I have also suffered drifting, and unpleasant jumps/hiccups.

Good tracking is not just about hardware, I suspect occulus had nice polished algorithms to filter out those kinds of anomalies from the tracking datas. Algorithms I suspect Pimax may have not invested much time into unfortunately, and the purpose of this thread was precisely to make some suggestions (at least to try to “solve” the wobbles, or if you prefer “make this LH tracking more tolerant” in case the cause could come from my environment or anything else in my setup).

Again, the thread wasn’t meant to compare both. You preferred to focus on this instead of answering/discussing the main subject of my post. Just look at your quote, from several paragraphs you have isolated the very few words that could let you answer off topic.

Now if comparing both is really the only thing you are interested in I can give you those other 2 arguments: IR cam are smaller, making them a bit less visible on your walls, but more importantly they are 100% silent and that would have been very nice for your off-ear DMAS ! :wink:

And what would have been the manufacturing cost of 3x IR cam compared to how much you pay to source 2x v2 lighthouses ? (I take those figures because I know from experience 3x IR cam is required for excellent IR tracking, and 2x LH was claimed by the LH users to be enough to do the same than 3x IR, but I’m not sure at all this is true).

For me, after having compared both in optimal quantity (3 for IR, 2 for LH, v1 and now v2) I don’t see where LH would win over IR cameras. The only advantage of LH is not taking usb ports on the PC (but takes power outlets instead) and maybe less CPU horsepower, but it didn’t seem that demanding when I was using 3 IR cams.

It is also possible my opinion is affected by polished oculus algorithms Pimax may not have. That would be a good news as this would at least leave a possibility the tracking with my 5K+ could get closer or even on par to what I have been experiencing with the CV1. As I have almost not used the OG vive I own, I’m curious to test how it handles LH tracking (never tried that before because I was focussing a lot on seated with racing games).

1 Like
6

Agreed. I’m wondering if you have defective hardware or shiny surfaces in the room or something weird about your setup. My base station tracking is rock-solid.

3 Likes
7

Yes, some thing is wrong, My tracking is pixel perfect.
P.S. try to get IR cams to work in the dark, LH work fine.

3 Likes
8

Don’t mix things up, IR cam is not the current inside-out with camera.
IR cam obviously has no problem at all with even 100% darkness.

9

I used CV1 with 3 sensors for 3 years without a hitch, not that I have had any issue with lighthouse.

10

Even if that was the result of a not optimized LH placement it would then be worth it to have a processing that makes it work perfect with that placement (mind we are not talking about canceling mirrors interferences).

If the tracking needs hours of hit and miss to find “the” correct LH placement + removing/covering any interfering object/surface (I’m not talking about occlusion, only reflections) then the tech is just bad/poor because it is too sensitive to the setup and environment, making excellent setup a real pain to achieve.

That’s what I mean when I say filtering-out the tracking anomalies would make the tracking system more tolerant. It would make the tracking setup (for flawless tracking) easier to the user.

I can’t accept your “you failed to place your LH” for the setup I’m using. I have a LH tightly fixed to the wall with the official mounts, good height, good angle, no mirror/window nearby or in the field of view of the LH, and also a LH just sitting on a bookshelve (or both of those LH simultaneously). “This” should just work.

IR cam didn’t require anything more than that and even for LH this should be a completely valid setup (= not failed one as you say).

I’m not saying I don’t trust you or other people claiming they manage to get “pixel perfect” LH tracking, but I’d like you make a video like mine then, HMD laying somewhere and a very close (few cm) virtual object showing it doesn’t move at all (+ moving controller in the background to prove not fake^^).

I have taken my OG vive out of its box and the result is the same (with 1 or 2 v1 LH).
So it could be something in my setup (no necessarily LH related, could be on the PC side), or just LH tracking tech itself.

I have done even more strict test:

I have used 2 big cardboard boxes to create a squared tunnel (approx 1.2m x 50cm) to isolate completely both the LH and HMD from any possible reflections interferences.

The wooble/shaking is still there. Interestingly it is there with the OG vive too but much less pronounced than with the 5K+.

Finally I have also tested the 5k+ in 9 axis mode (= no LH, only using internal gyro) and only with that I obtain a (almost) pixel perfect tracking like I would be so happy to have with 6DOF (LH).

Here are the video captures of the tests:

OG vive + 1x v1 LH fully enclosed in cardboard:

5K+ + 1x v2 LH fully enclosed in cardboard:

5K+ 9 axis mode:

It’s not just about negatives alone but weighting negatives VS positives (and for both tech). Now this is off topic anyway, I have LH tracking with my 5K+ and that won’t change, but that tracking is not as polished as I would like and this could change, and this is the main subject of this thread.

11

Do you have the synk cable ?
Can your LH see each other ?

12

The tests above are with single LH.
When testing with 2 v1 LH they were seeing each other.
When testing with v2 LH, this version don’t need to see each other anymore

13

wrong!!
CV1 tracking is really good, way much better than inside-out. I never had any issue since day one (4 cameras) and I still use it.
The ONLY reason they don’t use it anymore is… money!
cheaper headset = more customers = more benefits

14

This topic was automatically closed 60 days after the last reply. New replies are no longer allowed.