A little language.
A clearer picture.
Short definitions, concrete examples, and original references.
8 terms
Actuator
A component that turns supplied energy into physical motion or force. An electric motor driving a joint is one example.
In practice A humanoid’s knee actuator helps produce the force required to support and move the leg.
Honda’s November 2000 ASIMO announcement ↗Degrees of freedom
The number of independent coordinates needed to describe a system’s configuration. A joint count and a degrees-of-freedom count can describe different things.
In practice A simple hinge needs one angle to describe its position.
Modern Robotics, configuration and degrees of freedom ↗Reinforcement learning
A way to learn behavior through interaction, using a reward signal to improve expected cumulative results.
In practice A simulated controller can be rewarded for staying balanced while following a target motion.
Spinning Up, reinforcement learning concepts ↗Policy
A rule that maps what a system observes to the actions it takes. A policy can be learned or specified by a designer.
In practice A walking policy selects actions using information such as joint positions and a requested direction.
Spinning Up, reinforcement learning concepts ↗Sim-to-real
Transferring a method or behavior developed in simulation to physical hardware. Differences between the simulated and real systems can affect the result.
In practice The H2O study trains a motion controller in simulation before transferring it to a robot.
H2O research paper, March 7, 2024 ↗Teleoperation
Live operation of a machine from a distance. Software can assist with parts of the control while a person provides goals or movement input.
In practice A human motion capture system supplies targets to a robot controller.
H2O research paper, March 7, 2024 ↗Retargeting
Mapping movement from one body or model onto another with different proportions and joint limits.
In practice Human arm motion is converted into a reachable pose for a robot arm.
H2O research paper, March 7, 2024 ↗Vision-language-action model
A model that connects visual observations and language with action outputs. Its performance still depends on training, hardware, and the task conditions.
In practice Figure describes Helix as the vision-language-action system its Figure 03 hardware supports.
Figure 03 introduction, October 2025 ↗No matching terms. Try a shorter search.