This Coffee Table Knows “Come” Better Than My Puppy

Coffee tables are great, because they can hold beverages and tired feet equally well. But they have one glaring disadvantage, which is that they’re notoriously stationary and are there in the way, even when you aren’t drinking anything and don’t want to put your feet up. The only sensible solution is to build an ambulatory coffee table that comes when called, which is exactly what Concept Bytes did in their most recent video.

If a walking coffee table seems familiar to you, that’s probably because you saw one making the rounds on social media about eight months ago when Concept Bytes unveiled their first creation. But they’ve improved upon that original idea, rebuilt the entire thing, and given it some neat new capabilities. The most exciting of those is the ability to track humans and walk over to them when commanded to do so — something that even my puppy seems unwilling to comprehend.

Though it looks similar, this new coffee table isn’t the same as the old one. Concept Bytes redesigned it and rebuilt it. The table walks on 3D-printed legs based on Giliam de Carpentier’s Carpentopod linkage design. That was, in turn, inspired by Theo Jansen’s Strandbeests. The legs move so smoothly that they almost have a rolling contact point, like a wheel. That makes them very efficient and drivable with relatively low-power motors. An Arduino Nano RP2040 Connect controls those motors through H-bridge drivers and listens for the “come” command (a double clap) through its onboard microphone.

A second Arduino Nano RP2040 Connect in a remote control lets the owner move the table around with input from a joystick. The two Arduino boards communicate over WiFi and the table has tank-style steering, so it can move forward, backward, and turn according to that joystick.

For the coffee table to be able to come when called, it needed to be able to identify a human and to determine that human’s location. One’s first instinct might be to use something like computer vision for that task, but the Concept Bytes team found a far better solution: mmWave sensing. That is essentially radar based on a 24GHz (not 2.4GHz) radio signal, which is particularly good at differentiating between humans and everything else. It is also very precise within its operating range of around eight meters.

The sensor they chose is an AI-Thinker RD-03D, which is very affordable (about $13 USD) and communicates through UART. The table’s Arduino scans the room with that sensor and looks for people, so it can then walk to them when it hears the claps.

Now if only I could install an Arduino and an AI-Thinker RD-03D on my puppy, she might actually listen as well as this coffee table does.