Comment by NalNezumi

6 hours ago

As part of my thesis work almost 10 years ago I worked on a robot vacuum cleaner, (working on their sensor data) and one smart hardware implementation they had was that they had separated the computer vision module from the main board.

This way, only processed vision data would be physically sent to the main board. This constituted of mostly just "line segments", almost like a sparse point cloud, to detect obstacles and edges. They argued that this was more privacy safe because there's no way for the main module to access any raw vision data. It did however make the SLAM part harder to make work.

In hindsight, a good decision. I got one as a thank you for thesis work and it's still running just fine (with battery and brushes replaced once) and good to know that with the years of software update it still can't check me walking around in underwears in my apartment

I'm obviously wondering which company this is but understand the anonymity. I have a LIDAR vacuum to avoid cameras and from connecting to its debug socket could see the point cloud enough to know that it was very granular.

  • Which brand did you get? I’ve always been hesitant to get one, I’d like one that I could reduce the amount of data sent out to the cloud