I tried the above, and got to the following conclusions:
Faster leave state detection
I’ve set PREF_BEACON_EXPIRATION to 150 seconds. I could go lower, but then it starts to report false positives (on leaving). Just around 50% confidence it hits 100% again. This could still work with HA, but I’d like to keep MQTT traffic as low as possible considering I’ll probably be using this protocol a lot in the future.
Reducing noise
I’ve set PREF_FAIL_FILTER_MANUFACTURER_ARRIVE="Apple|Samsung|Google" and blacklisted a couple of my devices that do not meet this criteria. I feel the noise is a lot less indeed. But strangely it still allows 2 Samsung TV’s from my neighbour with manufacturer Samsung Electronics Co. Ltd.. The type is GENERIC_BEACON_PUBLIC, so shouldn’t these devices be ignored as well?
I’m curious how you did this. Are you running a script on your raspberry pi’s that listens to a MQTT topic, and you publish to that topic when HA restarts?
Now I’m noticing some odd behaviour. When I restart HA (which I do a lot) my home state defaults to not_home, though the confidence of both floors is 100. HA sees this as well since all messages are retained.
Restarting both monitor services it reports 100 confidence again, but HA isn’t changing the home state. When I actually move my Nut out of reach, as soon as it drops below 100 (and above 10) confidence HA succesfully changes the state to home.
As far as I know HA always changed the home state according to the confidence retreived, even directly after restarting. What am I messing up here?
edit: I can solve it through this automation, but I seriously doubt this was necessary before:
automation:
- alias: Get presence on HA restart
hide_entity: true
trigger:
- platform: homeassistant
event: start
action:
- service: device_tracker.see
data_template:
dev_id: rourke
location_name: >
{% if states('sensor.home_occupancy_confidence_rourke') | int > 10 %}
home
{% else %}
not_home
{% endif %}
source_type: bluetooth
I’m on my way out to work, so not a lot of time to look at this right now. Can you put the MAC addresses for your neighbor’s TVs in the blacklist file?
As far as the automation goes…I’m using the “not so binary presence detection” from here:
and so I have several automations on HA restart, for example, one of them is:
alias: 'Mark Anwen as home when HA starts'
trigger:
- platform: homeassistant
event: start
condition:
- condition: state
entity_id: device_tracker.anwen
state: 'home'
action:
- service: input_select.select_option
data:
entity_id: input_select.anwen_status_dropdown
option: Home
- service: input_boolean.turn_off
data:
entity_id: input_boolean.anwen_ext_away
Side note: I’m only running one instance of Monitor, because I don’t need to cover the whole house, just where we put our keys when we are home.
I did that, and it works. I hope it will ignore future devices from those manufacturers though.
That is quite interesting, although I’m not sure if I’m gonna need this. I have to think about it. But with the automation I’ve posted earlier I expect this runs the way I expect it to run. Although I’m not very happy with the not_home to home transition on every HA boot, since this sets some other automations in motion. I swear this wasn’t necessary before.
Now that I’m not tracking my phones anymore this is a valid point for me too. I might relocate my Pi Zero W to the garage. That way it detects coming home faster.
Thanks again by the way. Appreciate all your help.
I wasn’t aware of the restart function in monitor. How good of the creator to have this built in! But just to be sure: if you use -x all MQTT messages are retained. So doesn’t HA automatically get the latest correct data on a restart?
By the way, what do you mean with the ‘not-so-binary stuff’? I’m eager to learn.
By the “not so binary stuff” I meant the link I gave you earlier to Phil Hawthorne’s blog. There’s a section about “Stopping the Just Arrived State” that may or may not be helpful for your issue. I’m not sure why you are getting not_home on a restart - mine shows as unavailable in the history for a very brief time and then has the same status it had earlier. You can watch the MQTT messages in the MQTT section of Developer Tools in HA; maybe that would be helpful to diagnose the issue.
Regarding testing…when I was figuring all this out initially, I’d put my tracker in a small metal tin with a lid to make it be “not home”. It worked great - the tin blocked the BT signal, and I didn’t have to keep getting up from my desk to move my tracker out of range. As soon as I’d open the tin…boom! Home, confidence 100!
Nope. Tried once more & still could not see it. unfortunately I don’t have the time nor technical nohow to resolve this & for now have connected my work phone (Android).
With the persons I configured it kept resetting the state on HA reboots, so after a while working in HA and triggering some automations I got annoyed. I switched back to input_booleans since they remember states on HA reboot. It’s working as it should now.
Another problem arised. After 1 day I checked my MQTT sniffer for devices that have been detected (and so the messages retained) and for every floor that was more than 300! In just 1 day. So I think the solution is to somehow only scan for my preset Nut trackers, just haven’t discovered how yet. If anyone has a solution I’d love to hear it.
Did you get this solved? I have the same problem where script works for some hours and then just doesn’t detect my Samsung S7 anymore. I need to restart script to get it working again.
I’m a little confused when it comes to the regular addresses and the beacons. My wife and I have iPhones. They do respond to name requests so I have those addresses in the known static addresses, but do they also report as a beacon? If they do is there any advantage to figure out their beacon addresses and adding those as well?
Any idea what is wrong with my setup? I’ve got Pi Zero W and it’s detecting my Samsung S7 correctly for couple of hours. After a while it stops detecting the phone but the script is still running fine (I get OK when sending “monitor/scan/echo”). Sending arrival scan request from HA goes through to Pi but it still doesn’t find my phone. Even restarting the script by sending “monitor/scan/restart” doesn’t help, I need to reboot the whole Pi (power cycle) before it starts working again. I have also tried reinstalling the whole thing. Below is a part of log messages through ssh.
I’m having problems with the arrive/depart message, can someone please help me figure out what’s going on? I feel like I went through the entire thread twice by now
I’ve setup monitor on a Pi ZeroW (tried a 3B+, same behavior) and most of it seems to work fine. I can see messages through MQTT and I’m also able to add new MACs through monitor/setup/ADD STATIC DEVICE so MQTT is working both ways. Still, whatever I try, publishing to monitor/scan/depart and monitor/scan/arrive doesn’t seem to do anything! I’m looking at the monitor.sh bash script for logs and I’m aware that arrive and depart don’t get triggered if all known devices are home or away respectively. I have two phones added and one with BT on, the other off.
I don’t know where to look to further investigate. Other people seem to have had the some problem but the solutions posted in this thread haven’t helped me so far. I really hope this can work out, I had an early version of monitor working a year ago and it was just great for my situation. But because of happy family circumstances my home automation setup suffered from neglect and I’m basically starting from scratch again. Thx for any help!
Unfortunately no. Whatever I did it still resets to not_home on HA boot. In the end I reverted back to input_booleans which remember their state. It’s doesn’t feel like ‘the proper way’, but it works quite well.
I swear this wasn’t necessary before.
