Just commenting on the sleeping presence, if you look at some of the actual guides for placement of the 24GHz models, they show under or over a bed as an example, I’ve tried 60GHz and 24GHz on walls for an entire room that includes a bed and just like you, it fails to consistently register sleeping presence. But I’m sure under the bed it would work, especially the 60GHz models as they can pickup very subtle breathing and even heartrate. But neither can do this from an angle or at a distance. They must be very close and facing the human as if standing, hence placement under the bed looking up as in device mfgr examples.
You are right…it is vibration. This is just Physics and nothing can be done here. Everything is relative and in US, our houses are from lamber, so using mmWave sensor sensitive enough to pick breathing, will trigger floor vibration.
I used less sensitive type
It works ok for small rooms, like bathroom and toilets, but not for bigger room.
So, I am back to square 1. I can try to place one under the bed, but it should have companion (on the wall, or less sensitive in the ceiling)
I ordered DFRobot sensors, will play a bit, however, I do not have hope…again it is just Physics
Maybe you could use more sensors, by putting one in the ceiling of every room above until you reach the roof, and apply extra logic on each floor: something like “if someone is in the room above, ignore ceiling sensor”.
This would only help if the rooms above cannot be independently occupied, but for a one person home with the bedroom above the living room it could work ok, as the occupant can only be in one place?
I have a family and a giant dog (Grate Dane). And this is work around, not universal solution. I was thinking about one sensor on the wall and one under the bed. But again…this is brute force solution
There is a new device Theia that uses thermal imaging and radar to detect presence. It also uses BT Beacon to ID said presence. The device should be for sale by November. Senziio.com
The best way I tried is to build your own presence detection sensor based on esp32 board and ld2410 sensor. You can attach and light sensor to it also. There is a great esphome page for setting this thing up.
The downside is that you have to build it by yourself, buy all components, put everything together, adjust code to suit your needs and find or print case for it. But it is doable and those devices are much better then most you can buy of the shelf.
Well on the end of the day everything comes to a sensor that have to do all the work. I dont say that sensor you ordered is a bad one, I just want to say that, in my experience , if you want a good sensor then build it your self.
I may test some of these. But I currently use Thirdreality Zigbee motion detectors in all the rooms and they work great. For example, if motion starts turn on lights. If no motion in 5min AND TV not on, turn lights off.
I use the Bayesian integration to figure out which room I’m probably in, based on BT detection, motion, time of day, what devices are in use, etc.
It’s surprisingly accurate - for me, more so than any single type of device - and it costs nothing. You build your model around whatever devices you already have. Well… it costs quite a lot in terms of the time it takes to figure it out, but otherwise…
Thanks for this idea. Reading the Bayesian documentation it sounds like it would be a sensor per room. Therefore it could identify multiple rooms. Have you found a way to combine them to identify the room that a person is in? Or, am I overthinking this?
Bluetooth would be the way to do presence detection by person/device in my opinion.
There’s a Room-assistant addon and Bluetooth tracking integration I’m not familiar with
, but check out this post:
I had motion sensors in each room anyway, and I added Bluetooth proxies with the Bermuda integration to track phone and watch.
But there’s also time of day - 4am and I’m probably in the bedroom, and so on - and things like the Xbox - if it’s on, there’s a good chance I’m in the study. If you think about it, there’s quite a lot of data you can use.
I have a Bayesian calculation for each room giving the probability of it being occupied (probability is an attribute of the binary sensor), and a template to pick the room with the highest probability.
The thing is, it doesn’t have to be 100% for it to work. Quite often, I’ll be in the living room, say, watching TV, with my phone charging in the kitchen, the dog wandering around and the Xbox left on downloading a game. I’ll get something like:
Kitchen 52%
Living room 84%
Study 31%
The template will say (correctly) that I’m in the living room.
Thank you for the informaiton on the Bayesian integration. I read through the documentation.
I have not found however a situation where, when I know there are only two people in a home*, and motion is sensed in two different rooms within seconds of each other, to then automatically know to just turn off the lights in all of the other rooms. Given the WAF and other factors (such as a guest mode turned on which would have it ignore any such logic), it is unfortunate but custom programming is the best way in such a situation as there are too many custom considerations to take into account…
Since this is all done “on a shoestring budget”, all sensors are PIR and I am unable to differentiate between human and pet, erring on the side of caution is always best practice
