Indigo Domotics

I find this a really hard one to solve with high reliability. I'm trying to work out a scheme to switch on some hue lights when someone enters a bathroom via a short corridor.

1. The layout is such that the bathroom door need not be closed so although I could install a door sensor, an open door doesn't necessarily mean that the lights shouldn't be on.
2. Can't presume user will have an iPhone or fob of any sort.
3. I have two hue motion sensors that I can use. I thought Sensor A in the corridor entrance to the bathroom could trigger then Sensor B in the bathroom triggers so in theory this means that someone should have walked into the bathroom. The reverse sequence should trigger on OFF command as it then indicates they have walked out of the bathroom. But I'm concerned that this sequence may not be reliable eg say someone is in the bathroom, having triggered A and B, and another person enters the corridor and triggers A, then B would already be triggered (since there's someone in the bathroom) meaning that the OFF command would be triggered!
4. As a failsafe I'd want to have an auto-off after a certain amount of no motion in the bathroom but what happens when someone is motionless in the bath for 30 minutes?

All thoughts welcomed!

I use x minutes of motion sensor inactivity in bathrooms to switch off lights and extractor fans. This works well in bathrooms as you tend not to stand still for long periods of time.

My partner seems able to sit motionless for hours in there so it might be a case of making the in-bathroom sensor sensitive enough!

Anyone else have any ideas?

You could do the reverse:

• Entrance A gets triggers and then if B is also triggered then turn on the light for X minutes
• If B-> A is triggered, indicating a reverse of the process then turn off the light
• If motion on B during countdown of timer, extend timer
• If no motion on B within X minutes then blink the lights to indicate that the user needs to wave their arms or something or the light shuts off in Y more minutes

I use this in a bathroom where the light is set to go off after 5 minutes and at the 4 minute mark the lights blink to indicate 1 more minute remains and if motion is again detected then an extra 5 minutes is added.

Now that we finally have room-wide voice recognition working well, that's no longer the holy grail of home automation. My vote for the next one is reliable presence detection. People have for years tried to use Motion Sensors in a way you're describing with varying degrees of success, but in my experience nobody has ever made it work well enough for my tastes.

There are a few other things kicking around out there (that wifi-based one someone posted about for instance). I'm very interested to see how this shakes out over the next few years.

implant an RFID chip and have 100 sensors in your house ..

joesan wrote:

I find this a really hard one to solve with high reliability. I'm trying to work out a scheme to switch on some hue lights when someone enters a bathroom via a short corridor.

All thoughts welcomed!

Have you thought about adding a door sensor to the mix?
(This is how I deal with no motion in bathroom for long periods)

I have found that the bathroom door is normally only closed if someone is in bathroom - most of the other time it remains partially open.
Hence the sensing logic which has turned on the light by motion - only turns off the light if the door is open/If closed it starts its count-down again



Glenn

in europe /continent you close the bathroom door if nobody in inside ...

jay (support) wrote:

There are a few other things kicking around out there (that wifi-based one someone posted about for instance).

That was totally me that posted that! Sadly the feedback was not reassuring.

I think the next holy grail will need to be a marketable, easy-to-use way for basic users to represent their rooms as nodes.
-There would need to be links between nodes, and a way of identifying activity in each node.
-This allows for simple vectors of the occupants, or, more effective but harder to code, probability distributions.
-There would also need to be basic presuppositions:
—Nodes are separated by X
—People can only travel Y distance in a given time
—People can’t leave the house unless leaving from a room with an exit (recently instituted in kw123’s plugin).
-There also needs to be an error-catch for the model to walk back its estimations to a last known accurate appraisal of room occupancy. Historical data prior to an error catch (after a thousand or so) would be ripe for machine learning (though I’m clueless how a target function is represented, much less how to get it into Indigo).

C4W I think-> you have referenced your home-grown presence detection, it looks dope.

Sorry for the rant- this topic clearly overstimulates me


Sent from my iPhone using Tapatalk

I use a single Aeotec motion sensor in our Bathroom. It controls both lights and ventilation. The delay before switching off is set to 90 secs IIRC.
The lights are set so that they dim slowly when switched off, which gives anyone caught sitting too still an opportunity to wave a hand to switch them back on.
My wife accepted this solution/compromise so it must be good!
A second sensor directly behind or above the toilet would probably prevent even this happening but for me it’s not necessary.

Tim

I have notion sensor and strips door sensor.

Motion turns it on for x minutes, unless Door is closed.

If you’re “standing”, Motion is sufficient; Door is likely to be closed if you’re “sitting” or undressed.


Sent from my iPhone using Tapatalk Pro

Thanks to all for your input. There's some great ideas here.

As I mentioned, any solution predicated on the door being closed would not be reliable enough. (Although, if the door was closed and the sensors were triggered that would be a very clear condition.)

It seems like C4W's refinements would be the closest to what I need.

I agree with Jay that reliable presence detection would seem to the be the next holy grail of HA. It will be great to see this solved - looks like mundmc is on the way to solving it already!

joesan wrote:

looks like mundmc is on the way to solving it already!

Ha! Just hoping to inspire somebody who knows what they’re doing!


Sent from my iPhone using Tapatalk