Virtual reality is a slow-moving field in some respects. While a lot of focus is put on optical technologies and headsets, there’s a lot more involved when it comes to believably placing a human being in a virtual environment. So far, we’ve gotten a good start at the visuals and head tracking, but interaction technology is still lagging behind a lot. [Lucas] is working in just that area, iterating heavily on his homebrew VR gloves.
The gloves consists of potentiometers, fitted with spools and attached to the tip of each digit on a wearer’s hand by a string. As the user curls their fingers, the potentiometers turn and the position of the fingers can be measured. The potentiometers are all read via an Arduino, which communicates back to a PC via USB. A custom driver is then used to interact with Valve’s SteamVR software, allowing the glove to be used with a wide variety of existing software.
Thus far, the system is merely tracking finger position, but the spool and string based design is intended to support motors down the line for each finger to create resistance, so the user can gain the feeling of touching and interacting with virtual objects. The project has the potential to be a cheaper, more accessible alternative than current off-the-shelf solutions. We’ve seen other hand-tracking gloves before, too – though none that track the spread of a wearer’s hand as well as the finger extension. If you’re working on precisely that, please do drop us a line. Video after the break.
I was wondering why go through the effort instead of using existing kapton flex sensors, then I checked the prices on Sparkfun. Ohh, that’s why. They’re about an order of magnitude more expensive than I thought they were.
Maybe they could use whisker research to make a more affordable and versatile flex and touch sensor.
https://arstechnica.com/science/2021/04/the-secret-to-a-rats-sense-of-touch-its-all-in-how-the-whiskers-bend/
I have seen a few different ways using microphones elements or one way is to use barometric sensors encapsulated in silicone rubber or fiber optic looped back to a low resolution ccd to get deflection. I would think the issue for many designs would be calibration for each startup and also how close you want each finger to have dead reconning in space as a finger moves in more then one dof. I am guessing that the simplest way to progress this design is he should go for braking each tendon as that will give more force for the space and weight constraints.
So it’s $50 for a “SparkFun Qwiic Flex Glove Controller” which you would need 5 pcs for 10 fingers, that makes it $200 just for the sensor. You could also use the “Flex Sensor” itself sold for $9. I can’t imagine that [Lucas] comes out much cheaper printing a case for each retractor when time taken into account.
But I guess the retractor design feels more natural. And maybe there is potential for force feedback to the fingers?
If I am not mistaken, I believe SpectraSymbol holds a patent on the flex resistors and that there aren’t really any competing products to drive prices down into a more accessible range.
Iv been looking at this over the last few days, and something I feel that is worth pointing out is that tracking relies on the expensive vive trackers. Otherwise it only works as a fancy controller. If you have trackers, you’re in luck and it’s cheap. If you don’t, it’s of limited use until you pay the extra money for trackers. I don’t have trackers, so I probably won’t build it.
there’s other DIY tracking options done and some in the works, which could be used for cheap.
But apart from that from what I understood you should strap your existing controller to it, which is tracked ofc. Not optimal in the long run but works
Yeah, unfortunately doing tracking from scratch well is really really hard, so I don’t blame them for just focusing on the glove. Tracking is more like 2 years full time of work, which is more than your typical hobby project.
Can just wholeheartedly agree. Don’t know if you remember me, but like a year ago you told me exactly that and now I’m over a year into making my optical tracking system, and finally getting to something resembling a tracking system (still working on calibration accuracy rn, not an easy task to make user friendly).
Out of curiosity, since osvr is basically dead, were you guys going to put multiple ir cameras into the final version of hdk, or was it always planned to have the one camera and be a sit down system? Personally I don’t have tons of space to move around, so sitting head tracking was always going to be enough for me. Any thoughts on directions I could look for improving the tracking system as it is on HDK2? The drift and (on the later sdk) judder are the only problems I have. I’ve thought about adding a normal camera for added calibration. Perhaps an intel compute stick could do the added load of image processing if trained to find the right reference points? I’m out of my depth here, so feel free to shoot me down.
The Arduino Nano BLE and IOT boards have built-in IMUs that one could use for tracking. Not necessarily the best solution, but they’re a start at least. I used one for motion tracking on a headset to control a pan-tilt on a FPV RC car and it worked decently well.
Reliability and durability of the bend sensors are the biggest problems with commercial ones: we’ll ignore compatibility with the “new” consumer VR. (There are commercial ones with finger spread sensors, but same problem with bend sensor durability.) And actually the string and motor is a great way to do force feedback haptics: similar “capstans” are the way most commercial haptic controllers do their thing. Optical and other (eg capacitive, like index controllers) techniques are gaining popularity for just hand tracking, but they have no feedback potential… Very neat!
Holy cow, Lucas is super-crazy-smart. I agree, I have yet to be dazzled by VR. It is a slow-moving technology, at least for us mere mortals. But I still think the tech, even the most expensive, is still not all that impressive.