DIY Telepresence Robot: Part 1
I built a telepresence robot for about $400
Four years ago, I was travelling back and forth between the office in San Francisco, and my home in London. I developed a great distaste for travelling at that time, and grew acquainted with the differences and inefficiencies of remote working, especially for a manager who is remote to a team together in an office.
To help with this, I decided I needed an avatar that could replace me in the office. Especially one one that could loom ominously on people at their desks, as managers are wont to do. So I decided that I needed a telepresence robot — a robot I could remotely pilot and stream video to and from my desk.
Telepresence robots aren’t a new thing, a few companies make them, but they cost in the range of two to ten thousand dollars, not something one buys on a whim. I decided to build my own for a fraction of the price. This blog post is about that build, which I had taken photos of at the time but never shared, the eventual build ended up looking like this, but I’ll discuss some of the build and tech that went into it:
First, let’s talk about the common features of telepresence robots:
- They must be able to stream video/audio to and from the remote operator. Telepresence means the operator needs to have some “presence” in the room, and most agree that this means streaming their face and audio to the robot, and also seeing and hearing the room the robot is in. As such the robot must have a webcam, microphone, a display for video, and an internet connection.
- They must be able to move under the control of the remote operator. We’re making a telepresence robot, and not just a videoconferencing system, so the remote operator needs to be able to move the robot in a way that is useful for projecting presence. Usually this means being able to freely move around the office, which heavily implies being battery powered, and wifi. Though some sell simple “telepresence” robots that are no more than a tablet on a rotating stand that sits on a desk; I’d call this more of a fancy videoconferencing setup.
- They should be approximately human-height. Nobody’s going to take you seriously if you’re going to be video calling from an ankle-height robot scuttling around the office floor. A telepresence robot should, in my opinion, be approximately human-height so that you can converse with someone without them feeling like they’re talking to a pet cat. However, often a telepresence robot also needs to chat with people who are seated without seeming like it’s talking down on people, so a height that’s somewhere between where your head would be when seated and when standing turns out to be a reasonable compromise.
To fulfil the first requirement, it was immediately clear that I should use a tablet that can run an existing video call app, since I didn’t like the idea of having to code my own. Fortunately, I had an old windows tablet lying around that was perfect for the job.
The second requirement suggests that we need a whole remote control interface to send control signals from the remote operator to the robot. This will make up the bulk of the software and hardware complexity of the robot — I needed not only a remote interface, but also all the code that would allow me to control the various motors.
The third requirement says a lot about the physical design and construction of the robot. Making something ankle-height that can drive around a room is easy, but making something somewhat human-height that can still drive around makes things a lot more complex and expensive; how do you have a stable but tall structure without it toppling over easily, but keeping everything small and inexpensive? Probably something really heavy in the base like a bunch of batteries.
I’ll cover these in the next part
