Microduck - A small two-legged robot that you can train to learn new skills | Pollen Robotics

Microduck is a compact, open-source biped robot designed for interactive learning and reinforcement training. Standing at 25 cm, it is ready to play out of the box and can be taught new tricks by users. With a pre-order price of $399, it offers an engaging experience for both hobbyists and developers interested in robotics.

1. Open-source Platform

   Microduck is built on an open-source framework, allowing users to access and modify its software. This transparency fosters a community of developers who can contribute to its capabilities.

2. Reinforcement Learning

   Users can train Microduck using reinforcement learning techniques. This allows for personalized behavior training, making it adaptable to various tasks and commands.

3. Simulation Training

   Behaviors can be learned in a physics simulation environment before being deployed on the actual robot. This feature ensures that users can refine their training strategies without risking damage to the hardware.

4. Multiple Pre-trained Moves

   Microduck comes with seven pre-trained moves, including walking, sitting, and kicking. Users can also create and share their own behaviors with the community.

5. Customizable Appearance

   The robot is available in four different colorways, allowing users to choose a design that suits their personal style.

6. Community Engagement

   Users can join a community on Discord to share experiences, troubleshoot issues, and exchange training policies, enhancing the overall learning experience.

1. Familiarize yourself with Microduck's capabilities by reviewing the documentation available on GitHub.
2. Start training in the simulation environment to understand how reinforcement learning works before deploying behaviors on the physical robot.
3. Experiment with different training strategies to see how they affect Microduck's performance.
4. Engage with the community on Discord for tips and support from other users.

What is Microduck?

Microduck is a small, open-source biped robot that can be trained to perform various tasks using reinforcement learning techniques.

How does Microduck learn new tricks?

Users can train Microduck in a simulated environment, where it learns behaviors that can then be deployed on the real robot.

Can I modify the software of Microduck?

Yes, Microduck is open-source, allowing users to access, modify, and enhance its software as needed.

What comes in the Microduck package?

The package includes the robot, a battery, a USB-C cable, and a game controller for easy operation.

Is there a community for Microduck users?

Yes, users can join a Discord community to share experiences, troubleshoot, and exchange training policies.

Microduck is available for pre-order at a price of $399, excluding taxes and shipping. Pre-orders opened on August 27, 2026.

1. Microduck offers a unique and engaging way to learn about robotics and reinforcement learning, making it suitable for both beginners and experienced users.
2. The open-source nature of the robot encourages community involvement and continuous improvement of its capabilities.
3. While the robot is capable of performing several pre-trained moves, users may find limitations in its ability to execute more complex tasks without extensive training.
4. The reliance on simulation for training may require users to have a basic understanding of reinforcement learning principles to maximize the robot's potential.
5. Overall, Microduck presents a fun and educational tool for exploring robotics, but users should be prepared to invest time in training and community engagement to fully benefit from its features.