humanoidsdata.com

Search

Search datasets, articles, and glossary terms for humanoids and embodied AI.

← All glossary terms

Hardware & control

Actuator

An actuator is a robot’s power mechanism that converts electrical, hydraulic, pneumatic or other supplied energy into controlled mechanical force, torque or motion. In a humanoid, actuators drive joints, often through transmissions; they are distinct from the joints and sensors, and their arrangement need not correspond one-to-one with the robot’s degrees of freedom.

Also known as: Robot actuator, Actuators

Updated

From control signal to joint motion

ISO 8373:2021 defines an actuator by its energy-conversion role. On an electric humanoid, the practical joint unit may package a motor, drive electronics, transmission, brake and sensors, but those parts should not automatically be treated as one signal or component.

Software abstractions are narrower still. The ROS 2 actuator interface represents a one-degree-of-freedom actuator. Physical mechanisms can instead couple one actuator to several joints, use several actuators around one joint, or leave a freedom unactuated. Actuator count, joint count and degree-of-freedom count are therefore not interchangeable.

Why actuator design matters for humanoids

An actuator must produce motion within torque, speed, travel and thermal limits while carrying its own mass. Transmission choices also affect efficiency, reflected inertia, mechanical compliance and how readily an external push can move the joint. The MIT Underactuated Robotics notes show why available actuation and force limits determine which accelerations a robot can command, even when its kinematic model has many movable joints.

What training data should preserve

Actuation records should separate command values from measured output. Useful fields include the control mode, motor-side and joint-side position, velocity, current, estimated or sensed torque, gear ratio, limits, temperature, fault state and timestamps.

That distinction is hardware-specific: Franka’s ROS 2 interface exposes desired joint torque and measured link-side torque as separate interfaces. Without equivalent metadata, identical-looking action vectors can describe different inner control loops and cannot be compared safely across humanoids.

Sources