Suggestion for an electronic go board

[Javerberg]

I have an idea (hardly original) for how to build a cheap electronic go board. The setup consists of a standard go board, a smart phone app, and a small box containing a motorised laser pointer and a simple camera. The box should be placed high up, for example on a nearby bookshelf. The board is read by camera, the laser pointer is used to tell where to place the stones. The phone controls the box, and is either running an AI or is connected to a go server. All software and hardware designs should be open source. I hope that the hardware for the box will cost under 100 €.

What do you think? Sounds fun? I'm an experienced software developer, but a newbie when it comes to hardware. I'm not currently suggesting a community effort, just want to hear your what you think about the idea. All suggestions are welcome.

[bogiesan]

I wish you best of luck.

I'm assuming you've figured out which of the many x-y/pan-tilt/mirror-shutter systems is economical and readily available from maker suppliers. Other silly details such as making the system aware of where the edges of the board are located, possible distortion from projection distances and angles, how to differentiate between b/w stones if replaying an .sgf.

will be watching here and on reddit.

[Javerberg]

Thanks for your suggestion. I have checked a bit for a pan tilt system, but I have yet to find an affordable pan tilt with enough angular resolution. If the laser pointer is 3 meters away and you need to control the bright spot within +/- 2 mm you need a rather impressive angular resolution. If I can find something ready it's great, otherwise my plan is to build it myself with two small stepper motors, and gears to exchange speed for accuracy.

As you mentioned, reading the board from the camera is far from trivial. I actually suspect this will be the hardest part. My plan is to use traditional image processing to find the corners (also not easy, for the first version I will place markers on the board), then process the image and use machine learning. Will most likely need to get several cameras, lamps and boards to be able to produce training data easier.

[John Tilley]

This idea came up on Dec 30th 2020 with a video of a prototype.

I recently had some time to myself and was able to make a working prototype of a new type of go playing robot that uses computer vision to read a board and a laser attached to a couple of servos to point on any physical board where the computer wants to move next. You can see the robot in action here https://imgur.com/gallery/Q6CIae2. I haven't seen anyone else use this approach so I wanted to share in hopes of inspiring others to help push this concept further. This was a pretty easy build and all the materials together cost less than $60 so I could potentially see there is a future where something like this design can fulfill the dream of playing against online opponents on a real board. I open sourced the project and it can be found here https://github.com/diegocepedaw/lasergo

I liked the idea, but there is a major safety issue sitting opposite a laser that is pointed at the Go board in front of you, should an error in the setup or programming make the laser point at your face. For this reason I can't see any commercial manufacturer being interested as it's a potentially dangerous product.

The Chinese boards that use built in LED to show the intersection to play on are safe but expensive. I think you need to look at using these around the board.

John Tilley

[Galation]

I think that I am not the only one that would better prefer playing online games with his goban than with a screen.

I must say that I both remember:

-an APP system able to read a position from a picture (the system required the user to register manually the four corner positions and the calculates the prospettic of the board from that

-during an amateur contest I attended, they were testing a solution for registering games to .SGF through a camera monitoring the game

About the safety risk for using a laser, I think that it is only a problem of intensity and that solutions can be easily proposed to solve that.

The main point for me is the opportunity to use one's equipment instead of a full electronic goban.

Galation

[Javerberg]

Thank you for the link! Interesting. Seems at least some of the work is already done.

I have myself been a bit hesitant about the safety aspect of a laser, and tried to come up with alternative methods that's feasible for a DIY project. One such method is, as you mentioned, to put leds along the edge the board. I once had a chess computer that worked in that way.

However none of the methods I have found have been really satisfactory, so I got back to the laser pointer. As far as I understand (I have to research this more carefully before actually building anything) a red 5 milliwatt lasers is reasonable safe. If you get it in your eye its like looking directly at the sun, no fun but no long term damage if its very brief. Never really considered this as a commercial project, more like a fun hobby project. With good instructions it should be easy for interested persons to build their own device.

[Satorian]

I've wondered whether this could be done with a cheap pocket projector. You could then paint all kinds of information onto the playing surface and would not need any motors.

[Galation]

this could be done with a cheap pocket projector.

my guess.
I am a very unpractical person, so not able to put together a prototype, but I like where this tread is going.
I think that a possible easy and cheap apparatus is very near, nearer today than it was in the past.

Galation

[bogiesan]

The Chinese boards that use built in LED to show the intersection to play on are safe but expensive. I think you need to look at using these around the board.
John Tilley

Even this idea is hardly new. I had a Reversi/Othello set in 1990 that used LEDs to indicate the chip's moves. Buddy of mine had a chess game with green LEDs in each square that was clumsy but playable. And there's an automated chess thingy with an x-y magnetic arm under the table: https://squareoffnow.com

The two different AI go boards' integration technology and manufacturing processes have become economically viable in just the past few years. I believe at least one of these Chinese devices hit the market in 2018 so it is already ancient history, tech-wise.

Anyway ... Looking at the evolution of physical go equipment that is supported by an integrated engine, anyone wish to speculate where this could lead in a few more years? What is the best way to play against an engine or a remote opponent if you are using a real board and stones? Probably practical to assume at least one person is reaching into bowls and placing stones on the board.
I can see:
1. 360 unique RFI chips molded into heavy plastic stones.
2. Overhead projection with machine vision that detects invisible markings on the stones. Requires a stand and boom.
3. A robotic arm is a cool manifestation but remains impractical and the thing would be in the way.
4. Additional refinement of the current technologies to reduce costs and increase elegance.
5. Economic full-size, full-color iPad-like touch screen that is robust enough to both use and sense real glass or shell stones.
6. What else you got?

[bogiesan]

I liked the idea, but there is a major safety issue sitting opposite a laser that is pointed at the Go board in front of you, should an error in the setup or programming make the laser point at your face. For this reason I can't see any commercial manufacturer being interested as it's a potentially dangerous product.
John Tilley

Nothing risky about properly designed table top laser systems. Remember virtual keyboards?
https://www.electronicshub.org/best-laser-keyboards/

[Javerberg]

I can see:
1. 360 unique RFI chips molded into heavy plastic stones.
2. Overhead projection with machine vision that detects invisible markings on the stones. Requires a stand and boom.
3. A robotic arm is a cool manifestation but remains impractical and the thing would be in the way.
4. Additional refinement of the current technologies to reduce costs and increase elegance.
5. Economic full-size, full-color iPad-like touch screen that is robust enough to both use and sense real glass or shell stones.
6. What else you got?

Both input (reading the board) and output (tell what the remote player / ai did) can be done in a lot of different ways, but most of them requires a for the purpose built board. I would prefer to be able to take my existing equipment and just add a box for less than a 100 €.

[fireproof]

6. What else you got?

I'll be very interested to see augmented reality glasses/headgear used for playing games. Or reviewing, for that matter. Layering remote or AI data over a real board could be very useful.

[bogiesan]

6. What else you got?

I'll be very interested to see augmented reality glasses/headgear used for playing games. Or reviewing, for that matter. Layering remote or AI data over a real board could be very useful.

Nice, probably easier to design and get to work properly than all the other ideas. Hmmm, cinematic techniques could be used to animate the display with different colors, transparencies, influence clouds, and other visual effects to help illustrate aituati9ns, variations, threads.

Who’s working on an AR go simulator right now?