↓ Skip to main content

RoomSenseIQ Revisited

·404 words·2 mins
Gabriel Guzmán
Author
Gabriel Guzmán
A regular human. (he/him)

When I last wrote about the RoomSenseIQ presence and environment sensors, there were a few things that weren’t ideal. To recap:

  • The device always broadcasts it’s own internal wifi network with no way to turn it off.
  • The device causes my Denon Home 150 to crash and enter a reboot loop that requires a factory reset to fix.
  • If my wifi access point restarts, the device does not automatically reconnect, I need to manually reboot it.

Other than the Denon crashing issue, I was able to live with the rest, and still get some useable data from the sensors, though I mostly stopped using them since it became increasingly annoying having to reboot each one whenever the wifi restarted.

This particular product seems to be very lightly maintained, there hasn’t been any meaningful activity in their github repo for over a year. A few other people have fixed some bugs, but those fixes never made it into the upstream firmware.

I was procrastinating some other project when I decided to see if I could get hermes to fix the issues by just pointing it to my blog post and the firmware repository.

Long story short, it read my blog post, cloned the repo, setup the build environment, validated the firmware was building, found and fixed the errors, tested that they were fixed, built me a new firmware to try, and created 3 PRs to fix the issues upstream.

LLM transcript available if you’d like to see my prompts.

Once that was all working, I asked hermes if it were possible to generate a vizualization of the mmWave data that the sensors were providing. In my head I thought maybe it could build a cool 3d point cloud of what the sensors were seeing, but apparently that’s not how these mmWave sensors work. These sensors can only tell how far something is away from it and if it’s moving or not, they only operate in a single dimension. Hermes wasn’t deterred however, and still build me a pretty neat website for vizualizing the data in real time:

Screenshot of mmWave radar vizualization
dashboard

LLM transcript for that part available here

The ease with which it was able to do that did give me pause though, since it made me realize that these sensors are wide open, anyone on my home network can connect to them and pull data, or change their settings. Not ideal, but that’s a problem for another day.