Lighthouse positioning sensor

[Jan Ciger]

Hello, 

Someone could be interested in this gadget:

https://github.com/ashtuchkin/vive-diy-position-sensor

Basically, it is a 3Dof position sensor based on the Vive Lighthouse technology. It doesn't provide orientation and uses only 3 light sensors, so it needs 2 Lighthouses to be visible at all times. However, it is also much simpler than the official hw from HTC/Valve. Combine this with an IMU and sensor fusion and a full 6Dof could be possible (probably with extra sensors, though). 

Regards,

J.

[cr0sh]

Hey, thanks for posting this. A while back, I read about how the Vive position sensing worked, and wondered if we'd see any diy attempts. I decided to find out if anyone had tried to do the same with the lighthouses, and found this:

http://uploadvr.com/hidden-in-the-maker-faire-is-a-new-working-version-of-the-lighthouse-basestation/

Ha! I guess the first version of this system was a diy effort! Look at the pics: Some makerbeam, a couple of custom driven hard drive motors (kept in their frames),ir or red laser modules, a couple of ir detectors for laser timing, bits of perfboard, some 3d printed frames and a couple of laser line diffraction gratings. I imagine inside those tubes mounted to the motor spindles are mirrors to direct the laser beam at a right angle thru the grating.

This should be easy (?) to replicate. What's sad is that the guy and his device got snatched up by Vive before he put it out there to the VR community - if he even intended to. Did Vive patent this? I'm not sure, but if they did, someone could potentially challenge it. How?

Well, the system works virtually identically to Ivan Sutherland's Twinkle box positioning sensor system that was developed for his early "Sword of Damocles" VR/AR experiments. Check it out if you can find it (some digging should turn up a pdf scan of the old research paper on it - I have a copy somewhere, if anyone is interested).

I think the main difference between the two are that - iirc - Sutherland's system used visible non-laser light, and cds cells, and no sync light pulsing, plus the sweeping system may have been different, but the overall idea was the same otherwise from what I recall.

[Jan Ciger]

On Thu, Dec 1, 2016 at 3:55 PM, cr0sh <junk...@gmail.com> wrote:

Hey, thanks for posting this. A while back, I read about how the Vive position sensing worked, and wondered if we'd see any diy attempts. I decided to find out if anyone had tried to do the same with the lighthouses, and found this:

http://uploadvr.com/hidden-in-the-maker-faire-is-a-new-working-version-of-the-lighthouse-basestation/

Ha! I guess the first version of this system was a diy effort! Look at the pics: Some makerbeam, a couple of custom driven hard drive motors (kept in their frames),ir or red laser modules, a couple of ir detectors for laser timing, bits of perfboard, some 3d printed frames and a couple of laser line diffraction gratings. I imagine inside those tubes mounted to the motor spindles are mirrors to direct the laser beam at a right angle thru the grating.

Oh it has been even worse, just look at the pictures: 
http://makezine.com/2016/06/21/exclusive-see-the-secret-prototypes-we-found-in-valves-vr-lab/

I believe they don't use difraction gratings as such but directly a line laser module. 

This should be easy (?) to replicate. What's sad is that the guy and his device got snatched up by Vive before he put it out there to the VR community - if he even intended to. Did Vive patent this? I'm not sure, but if they did, someone could potentially challenge it. How?

Actually Allan Yates is Valve employee since quite a while and the Lighthouse tracking system has been developed at Valve, so he couldn't just "put it out there". It is Valve's technology, even though a very similar system has been developed earlier by Nikon I believe. 

I don't believe Valve has tried to patent it, they are actually actively encouraging people to build 3rdparty accessories for it. They have also started to license the design to manufacturers, so next year there could be 3rdparty devices that work with it available. 

Re DIY-ing - well, Allan has posted on Twitter and elsewhere some schematics of the early versions. The problem is not the principle how the system works but the electronic complexity (and firmware) you would need to get something comparable with the performance of the Vive sensors. The system is line of sight, so needs many photo sensors in order to avoid occlusions. That means a lot of analog circuitry to condition these signals (Valve has a custom chip for it now!). Then you must process these signals,measuring the time from the sync pulse. With the large number of photo sensors, there is no way around of having to use an FPGA. Then you will need a microcontroller to pull the data out of the FPGA, to handle the IMU and to fuse the data (unless the fusion is done on the host/PC). And the entire mess still needs an USB & radio interfaces, battery, battery management, etc. It is a lot of electrical engineering and would be a huge mechanical challenge to fit it into a small controller without resorting to stuff like flat flex cables or custom chips - not something a normal DIY-er has access to. So while the system is certainly amenable to some DIY efforts, I don't see people going crazy about doing that any time soon - the amount and types of the skills required is very difficult to find in a single person, notabene a hobbyist who is not an engineer. Lots of EE, FPGA and microcontroller programming, driver development and some heavy math required (especially if an IMU is to be integrated).

The article I have posted does away with a lot of this - but the result is a 3DOF only sensor, using only 3 channels and that works only when it is "upright" (because it doesn't cover the second hemisphere) and has limited accuracy (~1cm). Good enough for a drone landing, but not so much for a custom VR controller. 

Well, the system works virtually identically to Ivan Sutherland's Twinkle box positioning sensor system that was developed for his early "Sword of Damocles" VR/AR experiments. Check it out if you can find it (some digging should turn up a pdf scan of the old research paper on it - I have a copy somewhere, if anyone is interested).

I think the main difference between the two are that - iirc - Sutherland's system used visible non-laser light, and cds cells, and no sync light pulsing, plus the sweeping system may have been different, but the overall idea was the same otherwise from what I recall.

Sorta kinda, but the Lighthouse is a lot more complex system than that because it scans in two planes, can handle multiple beacons and even encode data on the light signal (both synchronization & beacon IDs). Then there is a very important part played by the IMUs in the controllers & the HMD, which is the way how they achieve the low latency.

Oliver Kreylos has a good analysis of how it works:

http://doc-ok.org/?p=1478

Regards,

Jan

[cr0sh]

Hmm - been a while since I looked, but the pdf of Sutherland's paper is on my websight, and it appears that it works on a different principle than the Vive lighthouse. My bad...

[cr0sh]

Update to my last update: At the end of the paper, under "Possible Future Developments", the author speculates on a system identical to the Vive lighthouse. The paper dates from 1974. The title of it is "Twinkle Box - A three-dimensional computer input device".

[Jan Ciger]

On 01/12/16 17:04, cr0sh wrote:
> Hmm - been a while since I looked, but the pdf of Sutherland's paper is on my websight, and it appears that it works on a different principle than the Vive lighthouse. My bad...
>

I have read the paper in detail now, having only skimmed through before.

For reference, the paper is here:
http://www.phoenixgarage.org/attachments/uploads/twinkle_box.pdf

Actually, the Twinkle Box is pretty much the Lighthouse, only in
reverse. They have put the sensors in the "lighthouse" where they are
sweeping the working volume and the moving object emits light. The
synchronization isn't described, but most likely was over a cable.

Valve's Lighthouse has the light emitters (lasers) swept in two planes
and the moving object measures the angle to the Lighthouse by measuring
the time from the sync pulse (IR LED flash).

The math for localizing the individual sensors is the same.

The Twinkle Box has the advantage that it would be a lot easier to
integrate into moving controllers, because only lights with some simple
electronics are needed (e.g. a ball like on a PS Move).

The advantage of the Lighthouse is that sweeping lasers is easier and
requires less complex optics than implementing what is essentially a ver
crude high speed camera sensor. Also having the light sensors in the
controller permits the IMU integration, which is the key to low latency.

I have checked the "Future work" part too - indeed, that pretty much
describes what Lighhouse or the Nikon's iGPS
(http://www.nikonmetrology.com/Products/Large-Volume-Applications/iGPS )
does. iGPS uses data from at least two beacons to localize a single
receiver, the Lighthouse can make do with a single beacon if multiple
sensors gets illuminated during the sweep.

Regards,

Jan

[Jan Ciger]

On Thu, Dec 1, 2016 at 5:13 PM, cr0sh <junk...@gmail.com> wrote:

Update to my last update: At the end of the paper, under "Possible Future Developments", the author speculates on a system identical to the Vive lighthouse. The paper dates from 1974. The title of it is "Twinkle Box - A three-dimensional computer input device".

There is now another thread on Twitter where Trammel Hudson (https://twitter.com/qrs) is building a Lighthouse tracked device using the official Valve sensor chips:

https://twitter.com/qrs/status/805193980203954176

https://twitter.com/qrs/status/805102989849612288

https://twitter.com/qrs/status/805498311411204097

Should give you an idea what is involved.

J.