Reference / 28 terms
Glossary
A working vocabulary for the machines, models, and methods that connect intelligence to action.
- Actuator
- A component that converts energy and control signals into physical motion.
- Autonomy
- A system’s ability to perceive, decide, and act without continuous human control.
- Behavior cloning
- Learning a policy by imitating demonstrations of desired behavior.
- Control policy
- A rule or learned model that maps observations or state estimates to actions.
- Degrees of freedom
- The independent ways a mechanism can move.
- Digital twin
- A computational representation of a physical system used for analysis or simulation.
- Embodied AI
- AI that learns or acts through a body situated in a physical or simulated environment.
- End effector
- The tool or device at the end of a robot arm, such as a gripper.
- Force control
- Control that regulates contact force rather than position alone.
- Foundation model
- A broadly trained model adapted to multiple downstream tasks.
- Grasp planning
- Computing how a robot should position and close a gripper to hold an object.
- Haptics
- Technologies for sensing or producing touch and force feedback.
- Humanoid robot
- A robot whose form or capabilities are organized around human-like interaction with environments.
- Inverse kinematics
- Calculating joint configurations needed to reach a target pose.
- Localization
- Estimating a robot’s position and orientation within an environment.
- Manipulation
- Robot interaction with objects through grasping, moving, or tool use.
- Motion planning
- Finding a feasible, collision-aware path from one robot configuration to another.
- Multimodal model
- A model that processes more than one data type, such as images, language, and actions.
- Perception
- Converting sensor data into useful estimates of objects, geometry, motion, or state.
- Physical AI
- AI systems that perceive, reason about, and act in the physical world through machines.
- Proprioception
- A robot’s sensing of its own joint positions, motion, and forces.
- Reinforcement learning
- Learning behavior through rewards received after actions.
- Robot foundation model
- A broadly trained model intended to support multiple robots, tasks, or environments.
- Sim-to-real
- Methods for transferring behavior learned in simulation to physical hardware.
- SLAM
- Simultaneous localization and mapping: building a map while estimating position within it.
- Teleoperation
- Remote human control of a robot, often used for operation or data collection.
- Trajectory
- A time-indexed sequence of positions, velocities, or controls.
- Vision-language-action model
- A model that connects visual and language inputs to robot actions.