Thanks, I have a few of those lying around so will give it a go!
I have around 16 devices ranging from simple ESP WROOM 32s that I got to use for ESPresence, to Shelly devices with bluetooth enabled and Everything Presence Lite presence sensors, which aare also proxies. The house has 8 main rooms and we’re never more than 3m away from a proxy. So once I’d calibrated everything it works really well. Placement and calibation is key, especially if you have a number of different devices as some appear to be better receivers than others. Despite my best efforts, we still get false readings. I put a proxy on the wall right behind the sofa we always sit on and often times a proxy in the room directly upstairs picks us up. Maybe I just need more 
Struggling here… I have 5 ESP32 D1 Mini and another 4 ESP32-C3 SuperMini.
Have set up the scale as well as done the Bermuda configuration.
My Phone with HA Companion is sitting about here physically around about 5 of above sensors (4 are ESP32 and 1 is a C3)
I am showing 5 BT Sensors showing relatively accurate distances as well
So why is my phone showing on the other side of the house?
I added more Bluetooth sensors. Now at 6 ESP32 D1 Mini and 5 ESP32 S3 SuperMini.
Max distance between any 2 devices is about 15 to 18 feet.
Do you think I need more?
It may be hard to tell on the image but the purple line is 10 feet.
Based on what I can tell you don’t need more (unless you live in a house with bomb safe walls).
I also have som positions in the house where coverage is less dense where I get som weird placement.
Are you sure your scales are correct?
If not: fix! 
If yes:
This is something I have planned adding in the function going forward…
If you take the 3 (or more) of the proxies that gives the lowest values and on an image of the floorplan draw cirkles around the proxies with the radius of the value given from the proxies.
Where these intersect (or there about) is where the phone should be illustrated by the red person icon in the image. It also highlights any issues you might have with settings of the Bermuda devices. They can be sensitive depending on placements and surroundings I have noticed.
Good luck and let me (and others) know what you found out!
1 Like
Thank you - will triple check my measurements. I do not have a blueprint unfortunately but fairly painstakingly created my drawing for my Floorplan. Should be within 1-3 inches at worst.
Are there any best practices on the following?
- Device used for setup - HACompanion vs BT Phone vs actual iBeacon?
- Chip direction - assuming using a flavor of ESP32, what orientation should the board be?
I’ve tried using my primary cellphone that has HA Companion app installed as well as my other cell that is just BT and my daughter’s iPhone.
For orientation, I have mostly put the ESP32 D1 vertical with the components facing away from the wall - same with the ESP32-C3. In places where I was able to mount high, I face the components away from the ceiling.
Good question but unfortunately I don’t have any answers. I personally use an Iphone and a apple watch.
I started to try to calibrate my proxies in Bermuda but could not really find the time and from the time I spent I did not get any wiser with values jumping all over the place.
I also just realized you are talking feet and inches. I’m not sure if this could be the issue but if you use metric (i.e. meters) instead, does this help?
Another question, if you only use the Living Room sensors is the precision good in that room. If there is no significant blocking of signal it should be rather precise.
I have to admit that initially I missed that measurements were in meters but caught that later the same day I installed. I have been converting mostly but acquired a metric tape measure later just to be sure
Unfortunately, no great luck for the Living Room either. I’m actually sitting there now yet at the red X. So about About 2.25 meters below where it is showing me.
bt_proxy_ls is an ESP32 D1 whereas the other 2 in the living room as C3. If I have any more D1 lying around, I may replace them as I am wondering if the 2 are not mixing well.
How important is line of sight. Most of my sensors are fairly revealed although due to the (low) height of the outlets where I have these plugged in, they are not always 100% clear LoS.
I am starting to think there is a different issue… Right now, BPS is showing my daughter’s phone in the hallway.
Problem is, she and her phone have been gone for almost 2 hours, roughly 15 miles away.
What could cause that?
I cleared the canvas and reselected her phone, same thing.
I refreshed the web page, same thing.
Cleared cache & cookies… same thing 
If you track a device in the graphical interface that is not there then it could generate some strange some strange results.
As a comparison, below is a illustration the kind of precision I get. This is not all over but it’s generally really good. When I enter zones with fewer proxies near by I get precision to match.
I see the front end part of the integration as a debug tool to visualize the precision and to see where more proxies need to be added. The point is that with the tracking in the back end we are able to do some really cool automations.
1 Like
I read your response again this morning and thought about it more… and ended up more confused about what BPS is targeted at accomplishing. And I mean absolutely no disrespect just trying to understand.
I was looking at BPS as a tool to give me visual representation as well as room specific automations.
Bermuda already tells me where/what room something is in. In fact, it tells me the correct room most of the time as opposed to BPS.
BPS on the other hand does not. It is at least 1 room away, if not 2. And not saying this isn’t a result of something on my part either although my floorplan is to scale and I have run through the config more than a few times.
Also, when I switch to a different phone to track - even if I Clear Canvas, refresh the BPS screen, or exit out of HA altogether - the newly selected phone is showing in the same place as the previously selected phone on the floorplan.
It almost seems like the orientation is off… Bermuda is showing fairly accurate measurements to 3 sensors [LW needs some tweaking] in the Living Room (lower portion of floorplan) yet BPS is showing me in the Sanctum (upper portion of the floorplan).
Bermuda vs BPS (this phone is in the Living Room where the X is)
Well, as I see it, it all depends on how you intend to use this integration. I see as part of further development to add a lovelace card to be able to track all devices in a graphical interface, however this is not top priority for me.
I use Home Assistant and automations extensively but I do not have an interface for it more than the standard. I want my “smart home” to stay hidden and “just do” things in the background. Bermuda is great, but I want for example to turn on the hockey game if I’m alone in the couch. Track me with my music when I walk around the house etc. Then I need a more precise position rather than just an approximate. I have a rather open floorplan house and I want to have more zones than Bermuda can provide by just trust the closest proxy. I also have cameras and have Double Take but realized that I needed more since this can be a rather dull instrument and I don’t have cameras everywhere.
My long-term thinking with this integration, rather than looking at a screen to see where the kids are, is that if we have precise tracking then we are able to hand over more of the control of the home to an AI.
I don’t have the issue with “orientation is off”. I just noticed that your sensor are registered in “ft”. What happens if you make sure that floorplan and sensors all are in metric?
2 Likes
Please see below images. I changed everything to Metric(Meters) for this example.
Room is 10 feet across, aka 3.05 meters. Phone is at edge of room.
I am starting to think it is either 1) bad mix of ESP32 D1 and ESP32 C3 or the orientation of the floorplan.
OK, after some “manual” debugging I see the issue…
According to your sensors you are 5.681m away from bt_proxy_hce but according to your drawing you are more 1 m. You are 6.162m away from btp_proxy_mc while actual distance is i would guess is more like 2m.
This is a problem when you have one senor per room (unless you have an open floorplan). Bermuda gives you the correct room but the distance to other sensors are off. If the distances are off then BPS will give you the wrong location.
I seem to have better precision in my proxies (don’t know why).
If you put three (3) proxies in e.g. Living Room will you get 1) the correct distances 2) a precise location?
2 Likes
I appreciate you continuing to assist!
I have 3 proxies in the Living Room but 2 are the ESP32 C3 which may be part of the problem… In watching the distances reported in Bermuda, the D1 measurements seem pretty stable from room to room but the C3 measurements vary much more and are inconsistent in reporting at all.
I believe I have a few more D1 units I can program and deploy to replace the 2 C3 in the Living Room (and else where if possible) to see if that helps reign in some of the distances being reported.
1 Like
Let me know your progress. I think this is good not only for BPS but also for Bermuda to see if we can find the most optimal boards for these kind of applications.
I see now that having the “circles” in the visual interface might be a good idea. Then it will be very clear that reported distances differ from the reported.
@agittins, if we add a “I am here” function in BPS, where a user manually place their physical location on the map, could that be an input to calibrate proxies in Bermuda? I’m thinking that we will with that be able to compare actual distance to reported.
1 Like
OK, some progress… two images below of 2 devices being tracked.
Both are in the correct room (yay!) and pretty close to their physical location.
I replaced the 2 ESP32-C3 in the living room so now there are 3 ESP32 D1 Minis for proxies there.
The bt-proxy-ls one is reading too short of a distance which leads me to another question. This proxy I put into a PETG 3D printed enclosure whereas the other 2 (bt-proxy-le2 and bt-proxy-lw2) are open/exposed.
Would that make any significant difference to BT readings? I would not have thought so. The enclosure has openings but…
You definitely have relative signal strength and reported distance - report them back to Bermuda with the station ID and the report - then math it backwards to the correct signal strength factors. Theoretically hitting each station with three pings from different locations (so triangulate with it and at least two other stations. - should be enough to calibrate the entire thing.
Does anyone have any thoughts, input, or suggestions on board orientation?
Perhaps just feedback on your trials?
For example, many of my proxies are plugged into wall outlets that are 10-12" above the floor. In these cases I have been trying to face the circuitry of the D1 Mini outward towards the room.
A few of my D1s are plugged into ceiling outlets and in these cases the circuitry is facing downward towards the floor.
The up-high and downward facing proxies seem to be providing the most consistent connections and often, most accurate data.
I know there are other variables but please share your experiences.
I know for a fact if I lay the one in my master bedroom on its back it doesn’t detect crap. 
1 Like