I use some precense decetion per room in order to controll the lights. This seems a threat to the WAF (Waife Acceptance Factor). Because my wife keeps turning off bluetooth and wifi. I fixed precense detection at home level by using a doorsensor at the front door (when you leave and lock the house, we use the front door). After experimenting with the app as tracker (that generates notifications on her phone and after a while gets disabled), I ended up with the network scanner integration looking for the mac adress (that I have configured to be stable in the phone).
However, for a room, this is not usefull, so I use PIR sensors (Zwave Fibaro and NeoCool) , combined PIR radar devices ( Zigbee Battery Powered Presence Sensor ZG-204ZM) and some esp32 based mmWave presence detectors. PIR sensors work most of the time, but seem not allways reliable. The mmwave radar detect someone sitting still, but not allways. (for example; I mounted a mmwave esp in front of me at my desk, to avoid the lights going off when I sitting still at the desk. I also noticed the quality of the devices may vary; some of the neo coolcams seems less stable than fibaro. Philio PST02B seems the most stable Pir sensor.
What are you experiences?
I have set the time that the sensor to report motion āonā for a pretty pretty short time (1 min) and used an automation triggered by the sensor āonā to start a timer. At the timer off event it gets triggered again and turns off the light. The time-out is dynamicly configurable this way. I can set the āonā time dynamically (shorter when at night or when a certain scene is active for example) and it is persistant when HA reboots.
Looking for suggestions to get this rock sollid and reliable enough to survive the WAF.
For example; is there a esp32 solution for using noise detection in addition? has anyone experience? (Ohh I do not install these google or other 3party mics in my home), just pure local stuff
This is probably not suitable for you as it tracks devices, which have to be left on. However on top of that I have set up a group of Bayesian sensors to tell me which room I am most likely to be in.
These use Bermuda data, but you can also take into account many more factors - time of day, day of the week, is the TV on etc. - to build up a picture of where everyone is, and you donāt necessarily need their cooperation.
You can also weight data to take into account the fact that Bluetooth may be switched off sometimes and phones left behind. Itās surprisingly accurate.
For home-level detection we use our iPhones with the HA companion app and find it to be pretty reliable. We also have an IKEA Somrig shortcut button by the front door though, which allows setting a āGuestā mode (so that lights donāt all turn off etc. when we leave the home radius), and also allows manual triggering of āAwayā mode and disabling smart doorbell person notifications for 1 min.
Weāre using IKEA Vallhorn (Zigbee) PIR motion/luminance sensors in most rooms. In some rooms these just trigger lights to come on, and in others where weāre not going to hang around without moving (e.g. utility room), lights turn off after a set time. Iāve found them to be most responsive when mounted above or next to a door where the wall meets the ceiling.
Iām also just starting to experiment with ESPHome devices with LD2410 radar sensors. We have one in the lounge itās working pretty well. Getting occasional false negatives (lights turning off when Iām sitting on the sofa very still). Also false positives but I donāt use it to trigger the lights turning on ā I use the PIR over the door for that. I probably need to keep tweaking the thresholds. Iām looking at deploying these more widely.
In the lounge Iām also using the state of our LG webOS TV to help i.e. donāt turn off the lights if the TV is playing.
I looked at Bermuda too, but my wife doesnāt always have her phone on her and our daughter is too young to have a phone yet, so I wanted a solution that didnāt rely on Bluetooth.
In terms of the āoffā trigger Iām using for like this, with input_number helpers to allow setting the time-out from the dashboard:
- trigger: state
entity_id:
- binary_sensor.utility_room_sensor_occupancy
to: "off"
for:
hours: 0
minutes: 0
seconds: "{{ states('input_number.utility_room_light_auto_off_delay') }}"
id: clear
The same automation is used for switching the lights on and off, with trigger IDs and āchooseā actions.
I am currently using Apollo Automationās MSR-2s in all our rooms for mmWave radar detection. I like these because they are very configurable. You can set thresholds for detecting movement and still presence at different distances from the detector. This is useful if you have noise in your environment, such as a ceiling fan.
Also, everything is available to Home Assistant through ESPHome, including all the thresholds. This allowed me to automatically change the thresholds when HA notices that the window fan is running in the room (using a power-monitoring smart switch) and to adjust the thresholds back when the fan turns off.
It took a while for me to tune these appropriately, but once tuned, theyāve been working very well. The one in our family room does a decent job of detecting presence even if weāre sitting still watching TV. However, the one in our bedroom does lose us when we are in bed for a while; this is due to a ceiling fan causing noise. Aside from tuning the thresholds correctly, I also tune the ātimeoutā or the amount of time between when the sensor loses presence and when it reports it. This has to be higher for rooms such as the family room where people are often still.
I have automations for all the rooms to manage the lights. Iāve also installed a little M5stack Atom Lite in each room that I use to control the automation for the room. It allows us to easily set the automation to full mode where lights are turned on and off, off-only mode where lights are only turned off, not on, and also to disable the automation entirely. I use the light on the Atom Lite to indicate the current state. We often need to turn the automations to off-only mode because we have cats that trigger the lights.
Iāve found the āstill occupancyā detection claimed by microwave detectors to be over exaggerated and even after tuning I found they false detect movement through walls or windows far too often. Gave up on them.
I use hardwired PIRs for 100% reliable room entry. Placement is key, you need to walk across the detection area not towards it. They donāt work reliably outside, neither do microwave or dual microwave and PIR sensors. For that I use on camera AI person and vehicle detection. This has been perfect.
Home occupancy detection is 100% reliable based on my security alarm state.
I donāt rely on room occupancy much for automation. Room detection for a few rooms is 99.9% reliable using Bermuda BLE and Private beacon tracking of my phone. I can also tell if my garbage and recycling bins are in or out on the street 100% reliably using Bermuda. However the more rooms I added to Bermuda the less reliable it became. Particularly because it cant distinguish between floors of the house. I can be right next to a beacon receiver upstairs and it will often report me as being in a room below. Having only a few rooms monitored at opposite ends of the house works very well though.
Firstly, it works best in a single-storey building as you might be closer to the device above/below you than the one in the same room. Either way, it will probably flick back and forth.
And secondly, I set it up because it was cool, not because I had a use case. The only useful function Iāve found is to send an announcement to the room Iām in, which is not often. I use PIR sensors to keep lights on etc.
I donāt see how that is going to work. If Iām in my lounge I can be up to 5m from the nearest beacon in the room. Straight down to the floor below could be as low as 3m.
Yes, floors in Bermuda does nothing more than expose an entity that shows the floor applicable to the currently detected āareaā - so thereās zero extra smarts, all it does is remove the need to create a helper for mapping the area to the floor.
