Updates

What has shipped for Microduck so far, from pre-orders to the open-source SDK and the simulator.

v2.0.0

Pre-orders are open

MicroduckPre-orderOpen source

Microduck in four colourways

Pre-orders for Microduck are open. The robot is $399 before taxes and shipping, paid once through Stripe, and it ships in four colourways: Cream, Graphite, Lavender and Sky.

What you can pre-order

  • Microduck - $399. The robot, a battery, a USB-C cable, a game controller and the seven trained moves.
  • Charger pack - $39. A dual charger and two extra batteries.
  • Accessory pack - $39. Laser pointer, NFC polaroid, two rollers, a ball and ten NFC tags.
  • Dev pack - $119. Spare motors and cables, batteries, a dual charger, NFC tags, a Hugging Face credit, a screwdriver and a screw pack.

Taxes and shipping are calculated at checkout and shown before payment. Nothing about the order is a subscription.

The robot that shipped with it

Microduck is 25 cm tall and about 800 g, with 15 motors, a camera, LiDAR and two IMUs, and a 50 Hz onboard policy loop. Seven moves come trained: standing up, sitting and standing, walking, turning, kicking, driving and grabbing.

How to order

  1. Sign in with Google so the order lands in your account.
  2. Pick the robot and any packs on the pre-order page.
  3. Pay once through Stripe. Taxes and shipping are added on the payment page.

The buy page has the buying FAQ, and the pre-order guide walks through the packs and the colourways in detail.

Questions about an order go to support@microduck.homes.

v1.0.0

Open-source SDK and simulator

MicroduckSDKSimulatorReinforcement learning

Microduck simulation and code

Microduck is open source, and that includes the parts that make it move. The Python SDK, the simulator and the reinforcement-learning stack behind the seven shipped moves are all public, so the robot you receive is one you can retrain rather than only control.

The parts that are open

  • Python SDK - connect to the robot, read its sensors, command individual joints, or hand control to a trained policy
  • Simulator - the same robot in simulation, for training and breaking policies without touching the hardware
  • Reinforcement learning stack - the training setup the shipped moves were produced with, ready for a reward function of your own

Seven moves, all retrainable

Standing up, sitting and standing, walking, turning, kicking, driving and grabbing arrive trained. Each one is a policy, so each one can be taken apart, retrained and replaced.

Where to start

Pre-orders are open on the pre-order page, with the buying FAQ on the buy page.

Microduck updates