Localize your cat at home with BLE beacon, ESP32s, and Machine Learning

I have a hard time describing it but the write up above triggered something.

I am interested in setting up wifi-valves on my home radiators - this would be really useful in a three storey house as getting balance right otherwise is impossible. But turning on lights and closing curtains just seems to be a foolish trade if complexity for convenience (ie we bought a big second hand car that had electric doors - they were great - till they stop working and you realise there are electric motors, sensors, relays and god knows what else where there used to be a lever.)

Anyway pretty much anyone can see that trade off (even if they do get excited by the electric doors and buy the darn thing anyway).

No the issue is the motion sensor and the lights. I have found myself in offices with those devices - it's a good idea saves electricity- and I understand how it works without any difficulty - and the solution is easy too - a motion sensor has timed out, therefore wave my arms around.

The interface is simple.

But when we add "smart" to it - when we write some algorithm to work out if a person is still in the room - either by direct sensing or cunning working out if they were in room A and now are in B, we end up in failure edge cases quite quickly.

And the problem is it is it simple or intuitive to work out how to "wave our arms" to fix it. If the room assistant is counting people in and out, what happens if I carry my daughter in, but she walks out - how many people are left in the room? How do I deregister ?

A similar thing happened with those electric doors - I let my kids jump out to go see Gran, parked too close to a tree. The electric door wanted to roll back to its full open position before coming forward to shut. But the tree won't let it go fully back. Ok the the engine back on and drive two feet - oh, I can't drive cos the door is open. Shut the door, but the door wants to go back first and hits the tree ... it took two grown men fifteen minutes to work that one out.

The problem is that software engineers like us have tried to make sensible defaults and decision trees - but we fail to cover everything.

Now FOSS is a partial solution to this - but for any appliance, for any IoT, for any thing that I am not paid professionally to understand, if I have to read the code to workout how to get around it, something is seriously wrong.

I am not sure where this leads - it's just ... user friendly usually is not. Simplicity is not just hard but sometimes orthogonal to sales and growth and, frankly I dont think most of our lives need more automation - i think we need help, actual assistance, advice and good curation.

I think I just wrote the opening paragraph on my chapter in MOOP.

Have never had a chance to try it as, of course, our cat refuses to say anything if a phone is pointed at him :-|

Turned out to be a coyote trap and there was a plastic link on the chain that attached it to the tree or whatever, that he had chewed through then limped home on three legs. If not for that link, he'd probably have died out there since the person who set it was clearly not checking it every 24 hours as required by law.

Luckily he was big enough that the trap couldn't close tightly enough to damage his leg and he was running around fine by the next day.

And if they are indoor make sure they have areas to run around, be sure to play with them (they are a pet! They are your responsibility!), and watch their diet (often people let them get too fat and it takes awhile for cats to lose weight).

catnip might introduce jitter.

Case in point: The vet. Our cat somehow knows from our behaviour that the visit is today, and disappears off the face off the earth. We had to cancel a visit because we spent more than two hours searching and were late.

So new appointment a day or 2 later, and we watch every step of the beastie like hawks. She disappears behind a cabinet, does not come out, and can be found neither behind or inside. Huh? It turns out the back of the cabinet is open, and there is a really tiny space between drawer and back. If I open a drawer, she goes up/down a layer.

I knew all the time she was less than 50cm from me, and it still took me 15 minutes to figure this out. I've been thinking about creating a tracker, but then , the cat healed enough to lower the frequency of vet visits.

So I’m wondering whether a timer that inflates a big red balloon after 4hrs wouldn’t be the best choice.

Looks like a classic case of the famous Jurissac Park quote: "Your scientists were so preoccupied with whether or not they could, they didn’t stop to think if they should."

The main difference between stats and ml is (i) community, (ii) stress on the details of compute and the focus on prediction accuracy rather than accurate recovery of parameters. A scheme that ensures good prediction even if the parameter is not recovered, for ML that's still a success.

One of my favorites was a color scan taken in multiple passes with filters. The cat was not stationary through the whole process

$ cat -L "en/us" /tmp/french-file.txt

It’s not wrong, and I would encourage the author to leave it as it is.

Why?

But my cat really doesn't like collars, so the real utility seems to be in locating where he managed to remove it...