Hugging Face unveiled the Microduck, a $399 open‑source duck‑like robot that can learn new tricks via reinforcement learning.
Hugging Face has entered the robotics market with the Microduck, a $399 open‑source duck‑shaped robot that learns tricks through reinforcement learning.
What Is the Microduck?
The Microduck is a compact, battery‑powered robot designed to resemble a cartoon duck. Its chassis houses a small camera, microphone, speaker, and a suite of sensors that enable it to perceive its environment and respond to user commands.
Open‑Source Hardware and Software
All mechanical drawings, PCB layouts, and firmware are released under permissive licenses, allowing hobbyists and developers to modify or extend the robot’s capabilities. The software stack is built on Hugging Face’s Transformers library, letting users fine‑tune language models that control the duck’s behavior.
Learning Tricks with Reinforcement Learning
Users can train the Microduck to perform new actions by defining reward functions and letting the robot explore via reinforcement learning. The robot records sensor data, evaluates outcomes, and updates its policy in real time, making it a hands‑on platform for AI education.
- Simple commands like "quack" or "turn left"
- Complex sequences such as obstacle courses
- Custom behaviors programmed via Python scripts
Pricing and Availability
Priced at $399, the Microduck ships worldwide with a starter kit that includes a charging dock, a set of interchangeable beak accessories, and a 2‑GB micro‑SD card preloaded with example projects.
Early reviewers praise the robot’s playful design and the transparency of its code, noting that it lowers the barrier for experimenting with reinforcement learning on physical hardware.
“Microduck makes AI research tangible for anyone with a desk and a curiosity,” said a beta tester on the Hugging Face forum.
For more details, see TechCrunch coverage of Hugging Face’s Microduck launch.