Bats are cheap and they eat bugs and talk about accuracy.
The Xandem system does look interesting but as you say, it is pricey. From the site video, it looks like it is close. I wish it were closer.
What I would like is a system that can decide that I am walking through a room to get to another room, and enable enough brightness to light my way. As I move toward a doorway, turn on the lights in the next room. If I stop inside the room, bring the lights up to allow the best visibility. If I sit in my easy chair to read, dim the lights in the room but adjust the light beside my chair for optimal reading levels.
I’m not sure PIRs and image recognition or Xandem can make this work???
not without a lot of pirs.
strategicly set distance sensors would be more helpfull there.
but you would need quite a lot from them also.
combining pir with distance sensors would be the best possible option for the moment i think.
together with a probablistic model
one way i have been thinking about is putting a IR measurement on both sides of a door
then when you walk in a room +1 room you left would be -1
Have no idea how to set it up think the logic make cents.
setups like this have been used before but it would still be best to include PIR in the mix as well. The problem with the idea you have is a thing called “tailgaiting” if 2 people pass through at the same time. Or a small child walks through, or your arm passes through before your body, etc, etc. Next thing you know everything is out of sync all over the place.
good Points
didn’t think that far ahead
that could be avoided with a distance sensor at the right height directly across the door.
if you combine that with a distance sensor that measures along the wall that has the door in it, you can tell exactly if someone leaves and how many.
I still seriously doubt that even with a ton of pirs or distance sensors you could get acceptable results for a single room let alone a whole house. Determining who is in a room and what lighting level was most appropriate for the activity would ultimately end up being a much larger project than HA itself.
i think you can get acceptable results with 3 distance for each door and 1 pir for each room.
the problem is more if you could place them on the place where they are needed, and if it wouldnt look bad 
the probabilistic model would be the most work, and give the most benifit.
with that you can do quite a lot.
with 1 PIR in my bedroom i can tell all the following (and actions are taken accordingly)
- if my wife goes to bed
- if i go to bed
- if my wife goes out of bed to pea
- if my wife goes out of bed in the morning
- if i go out of bed in the morning
- if there is a breakin
- if my wife is in the bedroom to clean or sort clothes
I guess I haven’t looked at the data stream from a pir. Are you saying the heatmap presented by a pir gives you enough data to determine activity?? and the individual in the room?
What pir are you using?
no probability is the keyword.
i just see the pir saying “on” and based on other settings i can tell those things.
off course it gets screwed up if i am ill and go to bed earlier then expected.
you just need a datasheet from most expected events.
my wife always goes to bed between 23:00 and 01:00 and i always between 2:00 and 4:00
if i go early i push a button on the dashboard telling HA its me and not my wife.
1 Like
You are relying on predictability rather than determing what is actually going on. After your post I went looking to see what I could find. This is an interesting read.
its a combination from whats going on and predictability.
and actually thats all we can do.
even if you have cams setup everywhere, you still need to have something that interpretes what the cam noticed. and that can only be done with predictability.
even facereqognition is about a certain chance that that face belongs to a certain person.
i think that datasheets with predictable things is still a way underestimated part from automations.
thx for the pdf. i have seen sortlike experiments from other places also (and this is familiar, so maybe i read about it before)
the biggest advance you can gain from predictability is that you can take action BEFORE things happen.
i use such things partially already.
my wife always makes sure that my disabled sister is good in bed.
my wife likes to have her heating blanket switched on when she comes to bed.
if motion is detected in the bedroom from my sis at a certain time, then the chance is high that my sis is goin to bed.
the chance is then also very high that my wife will go to bed in a short time after that, so her heating blanket is switched on.
off course that kind of things work best in predictable situations. some people live very unpredictble lives others very predictable. but even in the most unpredictable houses, there can be a lot of things predicted.
Doing a bunch of quick looking, microwave motion sensors look worth a look. They can work through walls, and quite inexpensive. Working through walls could help make direction and speed measurement possible. Turning lights on before someone enters a room. I will get a few and try them out.
i wish you luck.
i have seen videos from trying to do what you want with those, but it seems that they are not good enough (yet) togive you what you want. (but its a while ago i saw that so maybe the are better now)
nice, but i think thats not enough for whole rooms.
something i saw in other videos too.
in close range it works but covering a whole house is another story.
I realize it is about small movement and close range. Just that this is radar on a chip and an SDK. Think about all the sensors that go into self-driving or semi-autonomous cars. Those will become more affordable and easier to source. I likely won’t live long enough to see my idea come to be, but nice to think of the possibilities.
i think you will live long enough
every time i think something cant be done yet and i start creating another way to do it, its possible in a better way a half year later.
things we needed to create ourselve 2 years ago are now out of the box.
look how fast spoken commands have taken a place in automation.
the question we have to ask ourselves everytime is:
do i really want this now and put in a lot of efford and money in it, or do i wait a little and get it then because it was created by someone else.
1 Like
So my first attempt at remotely controlling the lights in my bedroom was a little more than 60 years ago. I fastened string to the light switch and looped it across the ceiling to my bed. Cool at the time. A lot has changed.
2 Likes