Compact biped form factor
Microduck is described as a 25 cm biped robot with articulated legs, head and neck, built for movement experiments rather than static demos.
Microduck
Microduck is a 25 cm open-source biped robot from Pollen Robotics for playful robotics and reinforcement-learning experiments. It can be used out of the box and retrained through a simulation-to-real workflow.
Overview
Microduck is a tiny open-source biped robot from Pollen Robotics, presented as a platform for reinforcement-learning experiments and playful robotics. The robot is described as 25 cm tall, with 15 motors, a camera, LiDAR, two IMUs, and a grasping beak.
The product is playable out of the box, but its main purpose is to let users train new behaviours in simulation and transfer them to the physical robot. The site describes a workflow that starts in physics simulation, deploys to the robot, and can then be refined and republished as new policies.
Features
On-board sensing
The site highlights a front camera, compact LiDAR, and two IMUs, one in the body and one in the head, to support perception and motion sensing.
Grasping beak
An articulated grasping beak lets the robot pick up and manipulate small objects, including the playful demo actions shown on the site.
Pretrained behaviours at launch
The robot ships with 7 trained moves in the box, and the homepage shows behaviours such as walking, sitting, kicking, grabbing, roller-skating, and getting back up after a fall.
Simulation to real workflow
The product page says behaviours are trained in physics simulation and then deployed to the real robot, with a refine-and-retrain loop supported by the open stack.
Connected onboard system
The press kit notes Wi-Fi and Bluetooth connectivity, a removable NP-F550 battery, and onboard compute based on a Rockchip RK3566 with AI accelerator.
Use cases
RL experimentation and sim-to-real transfer
Train walking, sitting, kicking, grabbing, or recovery behaviours in simulation, then deploy the resulting policy on the physical robot.
Out-of-the-box play and demos
Use the included controller and shipped behaviours to explore the robot without building software first, which is useful for demonstrations and early evaluation.
Behaviour prototyping and publishing
Iterate on behaviour policies, refine the simulation, and republish changes as new skills for the community or for a lab workflow.
Teaching and learning robotics
Use the robot as a small physical platform for robotics education, especially when the goal is to show locomotion, sensing, and policy training in a compact form.
Developer workflow on a local machine
Test the robot in a connected workflow using the open-source SDK, robotctl commands, and simulation tools on your own machine.
Pros and Cons
Pros
- Playable immediately with a game controller and autonomous behaviours included at launch.
- Open-source software stack covers the SDK, simulation, and RL training workflow.
- Supports a clear sim-to-real path for retraining and deploying policies on the robot.
- Includes practical sensing and interaction hardware such as a camera, LiDAR, two IMUs, and a grasping beak.
- Compact size and under-800 g weight make it a small desktop-scale robot rather than a large lab platform.
Cons
- The source says camera resolution, LiDAR range, radio versions, SDK languages, and age recommendation were still being finalised in the press-kit material.
- The software stack is open source, but the mechanical and electronic design files are not; the robot should not be treated as open-source hardware.
- Some details are only presented as introductory or provisional, including launch timing, first deliveries, and certain technical specifications.
FAQ
What is Microduck for?
Microduck is a 25 cm biped robot intended for playful learning and reinforcement-learning experimentation. The source says it is playable out of the box, and its open-source software stack lets you train new behaviours in simulation and run them on the robot.
How do you work with the software stack?
The source describes an open-source software stack, including the SDK, simulation environment, and RL training stack, and says the robot can be configured, updated, and monitored with robotctl commands. It also mentions training policies in MuJoCo and deploying them to the real robot.
Do you need to program it before using it?
The press kit says the robot ships with game controller support and autonomous behaviours at launch, so it can be used before writing code. It also notes that behaviours can be retrained on your own machine or on Hugging Face Jobs.
Is Microduck open source?
The source says the software stack is open source under Apache-2.0. It also says the mechanical and electronic design files are not open source, so the robot should not be described as open-source hardware.
What is the price and availability?
The site lists an introductory price of $399 before taxes and shipping, with pre-orders open from August 27, 2026 and first deliveries targeted before Christmas 2026.
Quick Facts
- Product type
- Biped robot
- Category
- Developer Tool
- Primary use
- Reinforcement learning and robot behavior training
- Platform
- Open-source software stack with simulation-to-real workflow
- Source domain
- pollen-robotics.com
- Price
- $399 introductory pre-order price before taxes and shipping
Alternatives à Microduck
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repositories and verifies changes with the real compiler, debugger, sanitizers, and tests before showing a diff. It offers a free tier plus paid plans, with editor connectors and zero-retention handling described in the source.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads inside Firecracker micro-VMs. It is designed for workflows that need machine-level isolation, private networking between sandboxes, and programmatic control through SDK, CLI, or MCP.
hob
hob is an independent workspace for coding agents that keeps agent sessions, terminals, history, and follow-up work organized around the tools and providers you already use. It is aimed at developers who want local control over routing, history, and workspace structure rather than a bundled model stack.
Manta AI
Manta AI is an autonomous web app testing tool for teams that want to map application behavior, catch regressions, and generate tests without writing scripts or maintaining selectors. It works from a URL and supports plain-English test flows, run results with screenshots, and scheduled or deployment-triggered checks.
SonOf
SonOf connects to your repo and PM tool, audits the codebase and surrounding product context, and turns approved work into shipped tickets with senior engineering review. It is aimed at founders and engineering leaders who need backlog help without hiring a full team immediately.
Ghost
Ghost est un assistant IA en terminal pour discuter, générer du code et lancer des tâches en ligne de commande. Modèles gratuits, Linux, macOS, Windows, open source.