The first humanoid actually deployed in warehouse work rather than filmed in a lab. Backward-bending knees, a torso that folds, and a design optimised for moving totes — not for looking human.

The arm that defined torque-controlled manipulation research. Seven DOF, joint torque sensors in every joint, and an interface that lets you command torques at 1 kHz — which is why it is in so many papers.

The most-installed collaborative arm on earth. Not the most dynamic or precise, but the ecosystem — grippers, vision, PLC integration, thousands of integrators — is unmatched.

The consumer quadruped that made legged robots normal. Its real achievement is the price: twelve QDD joint motors, a LiDAR, and a full locomotion controller for less than a used car.

The humanoid price collapse in one product. 23 to 43 DOF, 1.3 m tall, and it can stand back up after a fall. Under $20k for a walking humanoid was unthinkable two years before it shipped.

Built for oil rigs and tunnels, not demos. Sealed IP67 joints, a thermal camera, and a locomotion controller validated in explosive-atmosphere inspections.

The reference quadruped. Nearly a decade of locomotion research compressed into a product, with an SDK and payload ecosystem that actually works in industrial environments.

The $100 arm that changed who gets to do robot learning. A five-bar printed arm driven by six STS3215 serial servos, trained by teleoperation — you move it, it copies you, then it repeats. This is the hardware that made imitation learning a hobby.

A pair of arms: one leader you physically move by hand, one follower that mirrors it. Dynamixel XL430 servos give the leader real torque feedback, so teleoperation feels like the real thing instead of a mouse.

A full humanoid designed explicitly to be built by anyone with a 3D printer and a credit card. Printed cycloidal reducers, off-the-shelf BLDCs, and every CAD file published.

A small biped trained in simulation and transferred to hardware. Every part printed, the RL training pipeline published, and it actually balances — which is rare in hobby bipeds.

The most-documented open mobile robot base: a diff-drive chassis with an ESP32/Teensy micro-ROS bridge, a LiDAR, and a complete ROS 2 Nav2 stack that works out of the box.

Not a robot with a camera — a robot whose entire control loop (Betaflight at 8 kHz) is in your hands, with a physics-to-frame ratio no ground robot can match.

A 4-wheel skid-steer field rover with RTK GNSS, a soil sensor array and a camera mast. The archetype behind most student agri-robotics projects — and the one with the most real-world failure modes.

Two ViperX 300 arms on a sliding rail, four cameras, and a whole research programme behind it. The canonical platform for bimanual imitation learning — the thing papers mean when they say 'we evaluated on ALOHA'.

The reference ROS 2 teaching robot. Pi 4 + OAK-D + RPLIDAR A1 on a Create 3 base, with every tutorial in the ROS 2 ecosystem written against it.

The quadruped that set the template for every cheap QDD robot after it: 12 custom actuators, ~9 kg, and a gait controller that runs at hundreds of Hz. Its actuator design is the reason Robstride and friends exist.

A Roomba chassis with a ROS 2 node on board. Subtract the vacuum, add a payload bay and a documented interface, and you get the safest possible way to start mobile robotics.