Is there a better way to detect room presence?

Right now I am using pir motion sensors but they don’t work very good. The only other thing I have found is this http://www.xandem.com/ Is there anything else?

There are a couple of other similar devices. Most seem to be at the pre-order stage right now. There is also a solution from happy bubbles https://www.happybubbles.tech/presence/ that is more of a DIY home grown solution.

I did see that on BRUH Automation. Something I didn’t like about that system is the delay so you would have to us some combination of beacons and pir sensors. I also didn’t like that you had to carry the beacon on you and that might be ok for me but not my kids.

It would be cool if there was some software for ip cameras that would track faces or bodies like self driving cars do.

You could try Find (https://github.com/schollz/find), which uses the RF patterns detected by your smartphone and some machine learning to localise the phone indoors. It has support for HASS: https://www.internalpositioning.com/plugins/#home-assistant

2 Likes

I have the Happy Bubbles (don’t have any BLE transmitters yet, however - just ordered some) and will be using that for presence detection. For motion detection, I currently use PIR MySensors, but they often stop working, and are fairly inaccurate. So, I just ordered the Xandem kit (might be the first one on here?) and will post some thoughts on it when it comes in.

I’ll see if I can come up with a HASS plugin, as well, once I have it.

1 Like

Hey any update on Xandem? I’m interested in this system and would love to hear how it’s working out.

Omron D6T looks interesting too:

1 Like

Hey Exos,

Yes - I’ve had it for a while. I find it much better than the PIR sensors I had built (using MySensors), which ran into issues quite often, and I have found that hass and mysensors routinely have issues with each other - so it just wasn’t a great solution.

Xandem is a bit pricey, but worth it. I got the 10-node version. It’s literally a set-and-forget system. I setup a switch in hass that polls the Xandem API (local device API, very simple to use), and also push motion events to hass.

If you don’t mind spending the money and want something that works, get it.

Downsides: In a “real” home environment, it seems a little difficult to track where the person actually is. The “X” appearing and moving around the screen isn’t quite as accurate as they seem to make it look, but it’s not bad either. Also, no app yet that I know of. I also haven’t investigated to see if I can export that information to hass, so if you’re thinking room-based lights etc, that’s out. It also only shows motion - not presence. If you aren’t moving, then the system things you’re part of the scenery. It needs motion to understand what is different. The only way Xandem could make that different is you’d have to teach it a baseline, and if you ever left a bag of groceries, moved furniture, etc - then you’d have to re-teach it the baseline. That seems impractical, so the approach they used seems like a good solution.

I don’t have dogs or any other big pets (just rabbits), so I can’t speak to that either, but tweaking the sensitivity is needed, I had to drop it down a little even with just rabbits in their cages.

what do you mean by a “real” home environment? Do you mean multiple story house with pets?

My biggest concern with Xandem is being able to detect presence downstairs as well as upstairs and in the basement.

Yeah, ‘real’ as in not a staged house or one that was the system was developed and tested in. Ideally you don’t want to put the devices behind furniture etc, but sometimes you have to - for instance I have one behind a solid oak desk, TV, and a bunch of wires.

Pets depends on size, I think.

Regarding upstairs and downstairs, I really think you should just focus on deploying it only downstairs. It is not compatible for multi-level, which I could see being a problem on split-level houses. If you need multi-floor support, I think their answer now is to buy two systems - but I think they are working out how to support multiple floors. Right now, basically, they don’t support a z-axis.

Did anybody play around with those OMRON D6T sensors yet? (an also, in HA)
I’d like to use these for my future room detection, since I don’t want to rely on wireless technologies too much, as well as not having to carry a BLE beacon all the time…

If it was only a few bucks, I’d buy one and test myself, but it seems they are quite pricy

yeah they are quite expensive for something you must fiddle around with.

Another solution might be use the arduino thermal camera sensor:

seems it’s about the same pricing…
Distance should be up to 7 meters, so that would be good enough for standard rooms, once correctly calibrated and such, currently awaiting other stuff to arrive, this will be my next project

2 Likes

Has anyone here tried Aura?

1 Like

@Amanda_M Not available yet. Looks interesting though.

Aura does not seem intended for Room presence detection. I have also tried Happy Bubbles but could not get them to work reliably :frowning:

So for me this is still an unsolved problem!

I started a thread last year about presence awareness by face recognition. As @rb666 said in the previous reply this is still something that someone has to crack yet.

The OpenCV component works, but performance isn’t great on a Pi, and of course you have to hold still for the camera. Working on a better solution

What is needed is something akin to radar. Small devices in every room and an overall device to manage and analyze the results. Each family member would likely present a different radar signature. The analyzed data could show movement and predict destination and activity.
An alternative might be trained bats hanging in the corners of our rooms and finding some way to coordinate the squeaks and send the information to HA.

2 Likes

Bruh has a youtube video on a system similar to your radar suggestion - but it was damn expensive.
I personally think simple PIR sensors and a probabilistic model based on the layout of your home would be a good start, combined with image recognition