Collection · AI & Learning

AI & Learning Robots

Robots built around reinforcement learning and simulation-to-reality development

Featured Learning Platforms

Detailed profiles with specifications, pricing, and in-depth analysis.

1 Platform

Official demo

Microduck

Pollen Robotics (Hugging Face) · Compact Bipedal · Sim-to-Real

A 25 cm, 780–800 g duck-shaped bipedal robot with 15 motors, an 8 × 8 time-of-flight depth sensor, two IMUs and an articulated grasping beak. Its open-source software stack is built for training reinforcement-learning locomotion policies in NVIDIA Isaac Sim and Isaac Lab — including a demonstration across 4,096 parallel simulated environments — then deploying them to real hardware. Rockchip RK3566 with AI accelerator onboard, approximately one hour of runtime, and an introductory preorder price of US$399.

25 cm · 780–800 g15 MotorsIsaac Sim / Isaac LabUS$399 Preorder

Reinforcement Learning

Platforms designed so learned control policies — not hand-coded scripts — produce the robot's movement.

Simulation-to-Reality

Workflows that train a policy in a simulator such as Isaac Sim, then transfer it to physical hardware.

Open Development Stacks

Robots whose software is open source, so developers and educators can inspect and extend existing behaviors.

Accessible Price Points

Smaller-scale platforms that bring physical AI experimentation within reach of individuals and small teams.

Back to Robots Directory