• Welcome to the Speedsolving.com, home of the web's largest puzzle community!
    You are currently viewing our forum as a guest which gives you limited access to join discussions and access our other features.

    Registration is fast, simple and absolutely free so please, join our community of 40,000+ people from around the world today!

    If you are already a member, simply login to hide this message and begin participating in the community!

A completely different technological approach for smart cubes

Wish Lin

Member
Joined
Jun 18, 2019
Messages
673
Location
Taipei, Taiwan
WCA
2018LINW02
This is a concept I came up recently that just may be the future of smartcubes:
Exciting things this technology could/might achieve
  • The system can be installed to ALL STICKERLESS CUBES WITH A HOLLOW EDGE(Probably most people's mains by now), so you can make your main a smart cube!
  • A smartcube that doesn't need charging at all!
  • Considerably cheaper(I hope so....)than smart cubes nowadays like the Gan 356i or GoCube.
  • The system can (only theoretically)be implanted to all twisty puzzles(smart 4x4 ,smart 5x5, smart skewb, etc)!

Technical Details

This technology relies on passive RFID chips, which will be installed in the 12 edges of the cube(9 at least), and since the chips are passive, you don't need to charge the cube at all, it's just like a regular cube.

For anyone who wants to recommend alternatives, make sure the technology uses passive chips(if not,totally fine, the only limit is no battery) and has a HUGE signal strength gradient over a small distance. The bigger the gradient is, the better.

Current Design
This design is straightforward and should have a high success rate, but has the down side of squeezing your turning space into a box, and the box is harder to carry around. This particular design only works for 3x3.
box design.png
We have a box surrounding the cube, as shown in the picture, with the small spheres representing the 8 RFID emitters/receivers. Now, since the signal strength from the emitters will decrease when the distance between it and the chips increases(signal strength is inversely proportional to distance squared), we need each chip to return two key information to the receiver: the chip's ID number and the signal strength it gets from each particular emitter.

Let's consider only one chip for now. In total, each of the 8 receivers will get a value from the chip of it's own signal strength THAT IS DETECTED BY THE CHIP, in these eight values, there will be two value that are the highest(strongest signal), which will be the two closest to the edge, as shown in the picture, thus we can know where the edge is. (In the picture, the blue edge is closest to the two blue receivers and red edge closest to the red receivers, so the chip will detect the strongest signal from them)
box design 2.png
The same process is repeated on each of the 12 chips for every couple of milliseconds. When the cube turns, these values change, and by calculating these changes, we can locate each and every edge's movement, thus detecting the turns. 9 is the smallest amount of edge needed to have chips in them because for every move you make, at least one edge with a chip needs to be in that layer for the turn to be detected.(Tell me if I need to explain in detail)

Please post your thoughts about this below! Any technical thoughts(help!), usability problems and also, constructive criticism are very welcomed.
#This is still in prototype stage. Don't put too much hope on it#


This prototype is a smart 2x2 system that can be installed on ANY 2X2!! It is also super cheap(~$20)

(Simplified)Appearance:
overall look.png
The most important piece of technology in this prototype are Hall sensors: They are sensors that can sense changes in the magnetic field. On a L-shaped frame, there are 8 hall sensors arranged like this:

hall sensor pos.png
And the 2x2 will embedded with 24 magnets on every piece facing outward, like this(1 of the magnets):

magnet pos.png
The idea is simple: when you turn the cube, at least 2 of the 8 hall sensors will detect a difference in the magnetic field, thus we can detect turns.

There is still a small problem though: although the turns can be detected, but the direction cannot(i.e. it can't tell R from R'). I have a clever trick to solve this problem, but I don't want to publish it yet.



Rough Calculation for the price(This will get even lower if it ever gets mass produced):
  • 8 Hall sensors: 10¢ each
  • Arduino Nano: $4
  • Magnets: $1.5
  • 2x2(Meilong 2x2): $1.5
  • 3D printed Frame: Roughly $6 (will be close to zero if injection molded)
  • All the other stuff: $5
Total: $19(!!!!!!!!)
Check out the designs(up)

(6/6/2020) A brand new design on the way!
 
Last edited:

xyzzy

Member
Joined
Dec 24, 2015
Messages
2,876
I think 5 emitters will be plenty, but I am not sure AT ALL, since I know almost nothing about triangle location.
If you can (exactly) measure the distance of a chip from two emitters, that restricts its location to the intersection of two spheres, which is a circle. With three emitters, you can narrow that down to (at most) two points: one above the plate, one below. (If your emitters are exactly coplanar, then it can't possibly distinguish what's above the plate from what's below, but I expect this shouldn't be a problem if you just slap a "THIS SIDE UP" label on one of the sides.)

However, if all we know is the relative distance (something like "the distance to emitter A is half that to emitter B, but we don't know the exact distances"), then we need an extra emitter to handle the extra degree of freedom, so four emitters would be needed. Either way, five emitters is more than the bare minimum needed, and the redundancy could serve as a way of improving location accuracy.

See also: https://en.wikipedia.org/wiki/True_range_multilateration

Disclaimer: I also know nothing about the hardware required for this to work, so I might be completely misunderstanding how the RFID chips or whatever really work. Naïvely, I'd expect interference from the different emitters to be a significant problem.
 
Last edited:

Wish Lin

Member
Joined
Jun 18, 2019
Messages
673
Location
Taipei, Taiwan
WCA
2018LINW02
Disclaimer: I also know nothing about the hardware required for this to work, so I might be completely misunderstanding how the RFID chips or whatever really work. Naïvely, I'd expect interference from the different emitters to be a significant problem.
Yes, that could be a problem(thought I am also not sure) So I am thinking doing the emits/receives one by one between a small time interval to prevent them from interfering each other.
 

Wish Lin

Member
Joined
Jun 18, 2019
Messages
673
Location
Taipei, Taiwan
WCA
2018LINW02
Small update: The chips doesn’t necessarily need to be passive. They can be active as well, just replace the batteries with wireless charging coils(if possible)

Also, if the chip has a huge RESPONSE TIME DIFFERENCE when distance between it and the receiver changes, that should work as well.
 
Joined
Feb 23, 2019
Messages
649
Location
The FitnessGram Pacer Test is a multi stage...
Wait a second, is it a good idea to take this article down before I finish writing the patent? Seems like this can actually be a valuable technology judging from you guys’ replies.
dunno. I think of the cubing community as a community first and foremost. I wouldn’t worry about it too much but if you want to feel free
 

Wish Lin

Member
Joined
Jun 18, 2019
Messages
673
Location
Taipei, Taiwan
WCA
2018LINW02
Update:

1. I found that the idea of using RFID RSSI is not new to the world and has been developed under the name of "indoor RFID RSSI localization" for YEARS.
But still, The exact detail and technology varies quite a bit and I am the first one to put the idea together with speedcubing.

2. NEED SOME HELP HERE: I am having trouble finding which reader to buy that can return RSSI values. The ones I found now are either readers for 1m+ applications that has an RSSI accuracy of +-10cm, which is not going to work, or readers that cost $200 per reader. Please suggest me one if you somehow know one!

Impossible to use readers I found:
  • Mifare MFRC522 ------ No RSSI
  • Cottonwood reader -------- Expensive
  • AC3992 ------- Expensive
 

Wish Lin

Member
Joined
Jun 18, 2019
Messages
673
Location
Taipei, Taiwan
WCA
2018LINW02
Update: After checking a bunch of datasheets, I am confident to say that there is no way RSSI can be used in this project. Back to square one now, but I do have another idea. Will update after further research.
 

CodingCuber

Member
Joined
Nov 10, 2019
Messages
703
Location
Melbourne
WCA
2019HERR14
I absolutely love this idea. If any and every cube could be a smart cube this would help so much with practice and seeing how you do in your solves statistically. My question is though, would there be one particular piece of software for the technology or would different companies be able to get licenses or something to develop their own software. Love the idea and wish you the best
 

kubesolver

Premium Member
Joined
Nov 15, 2019
Messages
425
I think the only reliable solution to make any cube i-cube is to record a cube with a camera and write some software to convert video to a move sequence. This kind of problem of tracking fast moving items in real time is very similar to what full experience VR simulators do and the
solution is to have many high speed cameras and do image recognition.
 

Wish Lin

Member
Joined
Jun 18, 2019
Messages
673
Location
Taipei, Taiwan
WCA
2018LINW02
I think the only reliable solution to make any cube i-cube is to record a cube with a camera and write some software to convert video to a move sequence. This kind of problem of tracking fast moving items in real time is very similar to what full experience VR simulators do and the
solution is to have many high speed cameras and do image recognition.
Yes it is possible and is the only solution to literally anycube smartcube, but don’t forget that your hand will block L and R and parts of U,B,D and F, making the simple solution of N cameras pointing at N faces and analyze turns imposssible.

So the only possible solution is the computer will have to process it basically like a well trained human(Lazer0Monkey for instance) and as far as I know, such level of neural network for commercial use doesn’t exist.

(VR simulators don’t need to track a bunch of small pieces moving quickly in all three axis and its background is much, much easier to render away than some random guy’s desktop.)
 
Top