0

Initial planning -- signal triangulation?
Moderators: adafruit_support_bill, adafruit

Please be positive and constructive with your questions and comments.

Initial planning -- signal triangulation?

by nagstaku on Fri Jul 30, 2021 3:56 pm

Hello!

First, have to say, I recently listened to Limor on Hanselminutes podcast, and it was probably my favorite episode ever out of the 100 or so I've heard. It was very fun! AND, it reinvigorated me to kick off a project that I have long wanted to do, but I know I am going to need some help.

First, a little background on my skill level: I am a C# software developer and a devops engineer, so living on the top of the stack these days (for about 6 years now), but long ago in 2015 I got my first arduino and programmed it to sense soil moisture levels and water a garden for me. I am fairly comfortable reading documentation and I've lightly used breadboards and soldered things over the last 20 years.

My budget is hopefully somewhere between 100-800$ -- I've been wanting to build this for long enough now that I know I'm committed to making it happen. Once I have the bones built up, I believe I will build many projects from it.

Now, the project! I would essentially like to make a system that can tell me where something is (1-2ft proximity is fine), hence I believe signal triangulation is the direction I'm going? From there, I would like to build a matrix of nodes that can emit light (maybe every 10 feet). This should work in an outdoor space, filled sparsely with trees, with an initial testing scale of a 40x40ft area being fine, but eventually getting very big (200x1000ft?). Also, ideally the nodes don't have to be exactly placed in the grid, but might follow natural landscape features.

Suuuuper bonus points if we can get away with using wifi signal strength, and somehow use a cell phone as the 'emitter'.

The end goal will be that I can walk into an area, and have the lights within 10-15ft of me slowly fade on as I walk into the light-gridded area. I have included an obligatory mspaint photo to hopefully solidify what I'm describing.

So, if anyone has any documentation, products i should check out, other projects that have been done I might want to see, PLEASE SEND THEM! I appreciate any and all help I can get.

Thanks!
Attachments
image_2021-07-30_124908.png
image_2021-07-30_124908.png (33.44 KiB) Viewed 160 times

nagstaku
 
Posts: 3
Joined: Fri Jul 30, 2021 3:36 pm

Re: Initial planning -- signal triangulation?

by sj_remington on Fri Jul 30, 2021 10:25 pm

Pozyx works, and the company sells experimenter ("Creator") kits: https://pozyx.io/

RTK GPS works too, outdoors with a clear view of the sky. You need only one rover and one base station to get 10 cm accuracy or better. I recommend the C94-M8P (the USD $399 kit includes two identical modules, one to act as rover, one as the base station): https://www.u-blox.com/en/product/c94-m8p

if we can get away with using wifi signal strength
No chance of that. Radio signal strength varies with too many factors to be of much use in distance reckoning or localization. If it did work, you would have been able to buy dirt cheap localization modules on eBay 15 to 20 years ago.

sj_remington
 
Posts: 124
Joined: Mon Jul 27, 2020 4:51 pm

Re: Initial planning -- signal triangulation?

by nagstaku on Sat Jul 31, 2021 8:54 pm

Hey, so someone did share this, which seems promising for using wifi signal strength actually, and the idea of using this ESP32-S2-WROVER Module for Access Points (which are cheap at 4$ each)

https://eloquentarduino.github.io/2019/ ... n-arduino/
https://www.adafruit.com/product/4653

I'll probably research using the ESP32 first.

Admittedly, that pozyx system seems perfect, but the 700$ price tag for the lite kit is a slightly larger barrier to entry, but I miiiight still be interested depending on success of above.

nagstaku
 
Posts: 3
Joined: Fri Jul 30, 2021 3:36 pm

Re: Initial planning -- signal triangulation?

by adafruit_support_mike on Mon Aug 02, 2021 1:03 am

Signal strength just doesn't work. It's comparable to making a clock that tells you the time, to the second, based on how bright it is outside. Even assuming perfect repetition, no shadows, and all the other problems hand-waved away, what's the difference in luminous density between 11:59:58 and 11:59:59?

Adding the real problems like annual variations, clouds, and all the other real world details makes the job.. at least highly difficult.

For RF-based ranging, you need synchronized transmitters pumping out a repeating pattern of bits that, when compared to a delayed copy of itself, lets you know how big the delay was. To get 1' resolution you need timing accuracy of 1ns (the time it takes light to move 1 foot). That puts your frequency requirements up in the 10GHz range.

Honestly, for room-scale ranging based on a transmitted signal, it's better to work with ultrasonic audio. Sound moves about 1m per millisecond, so a 1MHz microcontroller can give you resolution down to the millimeter without working too hard.

Another option, which would be more effective in general and suitably magical, would be to use a camera and machine learning to identify people and their positions in space. Three cameras would give you image triangulation, and the algorithm would basically compare the current frame to a baseline image of the vacant space.

That system would be able to track the general location of everyone in the space at the same time, and could map occupied areas to available lighting, time of day, etc.

adafruit_support_mike
 
Posts: 63679
Joined: Thu Feb 11, 2010 2:51 pm

Re: Initial planning -- signal triangulation?

by nagstaku on Tue Aug 03, 2021 2:24 pm

Thanks for the thorough reply Mike

The camera option was actually something I had been curious about. I think the Jetson Nano board has some interesting projects out there, and work I put into image based ML systems could one day assist some other hair-brained ideas I have.

Do you know of any good intro projects on the subject that I could take the time to checkout and learn?

nagstaku
 
Posts: 3
Joined: Fri Jul 30, 2021 3:36 pm

Re: Initial planning -- signal triangulation?

by adafruit_support_mike on Tue Aug 03, 2021 5:17 pm

I’m afraid I don’t know of any such projects offhand. Every machine learning system is its own little world, and takes some time to learn.

For image recognition, the OpenCV is common and seems good: https://opencv.org/

adafruit_support_mike
 
Posts: 63679
Joined: Thu Feb 11, 2010 2:51 pm

Re: Initial planning -- signal triangulation?

by AgrimagWeb on Thu Aug 05, 2021 8:11 am

My opinion regarding the mechanism used in https://eloquentarduino.github.io/2019/12/wifi-indoor-positioning-on-arduino/: the perspective there is the opposite of what you have here. In that post, you have a "sensing" device that locates itself based on nearby hotspots and takes action based on that. Here, the ones that need to take action are the hotspots, which need to sense when you're nearby.

I don't think this kind of approach will work.

Other approaches should be more suitable: the most immediate I can think of (never used so cannot assure it will work) implies using an RFID tag on you and attaching a UHF RFID receiver on each node. When a receiver senses the tag, light up the LEDs. Don't know how much power these RFID receivers require, don't know how much they will cost (I see on Google they're ~ 200 $ each).

You have to investigate which RF technology covers your required range of 3-5 meters.

AgrimagWeb
 
Posts: 3
Joined: Thu May 21, 2020 4:26 am

Please be positive and constructive with your questions and comments.