Microduck

Microduck is a compact open-source biped robot designed for physical AI, reinforcement learning, robotics education, and experimentation. It combines sensors, programmable movement, simulation tools, and sim-to-real training for developing new robotic behaviors.
Pricing Model: Paid
https://pollen-robotics.com/microduck/
Release Date: 27/08/2026

Microduck Features:

  • 15-motor biped robot designed for programmable whole-body movement and physical AI experimentation.
  • 25 cm compact form factor weighing under 800 grams for desktop and classroom robotics projects.
  • Front camera, compact 8×8 ToF LiDAR, and two IMUs provide vision and motion sensing.
  • Articulated grasping beak allows the robot to pick up and carry small objects.
  • Open-source SDK provides tools for controlling the robot and developing custom behaviors.
  • Simulation environments support reinforcement learning before deploying behaviors to physical hardware.
  • Sim-to-real workflow allows trained policies to move from simulation onto the real robot.
  • Built-in behaviors include walking, sitting, standing, kicking, grabbing, roller skating, and self-recovery.
  • Wi-Fi and Bluetooth connectivity support communication and development workflows.
  • Game controller and ready-to-use behaviors allow Microduck to be operated before writing custom code.

Microduck Description:

Microduck is a compact biped robot from Pollen Robotics designed for physical AI, reinforcement learning, robotics education, and hands-on experimentation. The 25 cm robot combines programmable movement with built-in sensors and an open-source software stack, allowing users to explore how artificial intelligence can control physical robotic behavior.

The robot uses 15 motors across its articulated legs, head, and neck to produce whole-body movement. It includes a front camera, compact 8×8 time-of-flight LiDAR, two IMUs, microphones, a speaker, and an articulated grasping beak. These components allow Microduck to perceive its surroundings, move through different behaviors, interact with objects, and recover from many common fall positions.

Microduck can be used immediately with its included game controller and built-in behaviors. The robot comes with seven trained movements covering actions such as walking, sitting, standing, kicking, grabbing, roller skating, and getting back up. This provides an accessible starting point for users who want to experiment with the robot before developing custom behaviors.

The software side of Microduck focuses on reinforcement learning and sim-to-real robotics. Developers can use the open-source SDK, simulation environment, and reinforcement learning training stack to create and train new behaviors. A behavior can be developed in simulation and then transferred to the physical robot, allowing developers to test how learned movement performs on real hardware and continue refining the behavior.

Microduck is designed for developers, robotics students, educators, researchers, and makers interested in physical AI and robot learning. Its compact size provides a practical platform for experimenting with locomotion, object interaction, reinforcement learning, and multi-robot projects.

Microduck is available as a paid hardware product. The introductory pre-order price is $399 before taxes and shipping, with the robot available in four colorways. First deliveries are targeted for before Christmas 2026 in North America, Europe, and the United Kingdom.

Frequently Asked Questions

What is Microduck?

Microduck is a compact biped robot designed for physical AI, reinforcement learning, and robotics experimentation. It combines 15 motors, cameras, LiDAR, motion sensors, an articulated beak, and an open-source software stack for developing and testing robotic behaviors.

What can Microduck do?

Microduck can walk, sit, stand, crouch, kick, grab objects, roller-skate, and recover from many common falls. It also allows developers to train and deploy new behaviors using reinforcement learning and simulation.

Can Microduck learn new behaviors?

Yes, Microduck can run new behaviors trained through reinforcement learning. Developers can train policies in simulation, transfer them to the physical robot using a sim-to-real workflow, and retrain or modify them as needed.

Is Microduck open source?

Microduck has an open-source software stack covering its SDK, simulation environment, and reinforcement learning training tools. The robot’s mechanical and electronic design files are not included in the open-source offering.

How much does Microduck cost?

Microduck has an introductory pre-order price of $399 before taxes and shipping. Optional accessories are available separately, including a Charger Pack, Dev Pack, and Accessory Pack.

What sensors does Microduck have?

Microduck includes a front camera, an 8×8 time-of-flight LiDAR sensor, and two IMUs for sensing its environment and movement. It also has microphones, a speaker, and two NFC antennas for additional interaction capabilities.

Does Microduck work without programming?

Yes, Microduck can be played and controlled without writing code using its included game controller and built-in behaviors. Developers can later use the SDK and reinforcement learning tools to create and deploy custom behaviors.

Who is Microduck designed for?

Microduck is designed for developers, robotics learners, educators, researchers, and makers interested in physical AI and reinforcement learning. Its combination of ready-to-use behaviors, simulation, open-source software, and real-world robot hardware supports practical robotics experimentation.

Real User Reviews and Rating of Microduck

0
0 out of 5 stars (based on 0 reviews)
Excellent
Very good
Average
Poor
Terrible

There are no reviews yet. Be the first one to write one.

Share Your Experience:

Alternative to Microduck

Showcase your AI Tool – Add it to our directory today.