Day 0
Which robot?
The plan started as a 70 cm brushless walking droid. A day of reading later it was a 25 cm printed duck. Here is what we looked at, and what each one would have cost.
The goal is a robot that potters round the house, chats, and finds things. The first idea was a walking droid in the style of Disney's BD-X robots: brushless motors, a round screen for a face, a camera and an AI voice.
The options
- A full-size brushless biped. The open BDX-R design was the reference: 16 Robstride motors, about 19 kg and roughly 65 to 70 cm tall. The motors alone came to about £2,500 to £3,500, and nearer £4,000 all in. At 48 V with 42 Nm joints it's a real injury risk for a first walking robot, so it would need an emergency stop and a gantry from day one.
- A robot on wheels. Statically stable, with lidar for finding its way round, a tall neck for the camera, and a charging dock. A better fit for "wander the house and find things", at roughly £700 to £1,100. Less charm, though.
- Pollen Robotics' Microduck. A 25 cm, 800 g duck with fifteen servos, trained walking moves, a camera and a speaker, for about $399. Brilliant, but it's a product to buy rather than build, with no CAD published.
- Luwu Dynamics' XGO-Duck. A 3D-printable version of Microduck that runs on an Arduino UNO Q. The STLs, the board design and the training code are all public, the walking policies come ready-trained, and the estimate was about $400.
Why the duck
It's small enough to be safe on a desk, cheap enough to try, and it uses the same ideas as the big robots: bus servos, a walking policy trained in simulation, and a head that can be made to see and talk. The UNO Qs were already on their way. And it's a duck.
It was also clearly early. The project's own roadmap said no independent build had been completed, and the battery, print profiles and licence were still being settled. That's a risk, and also the reason this diary exists.
The servos went on order the next day: Day 1.