Robotics Glossary

55 essential robotics terms and concepts, from actuators to XACRO. Each definition links to the lesson where the concept is taught in depth.

A B C D E F G H I K L M N O P Q R S T U V W X Y

A

A* (A-Star): A graph search algorithm that finds the shortest path between two points by combining the actual cost from the start with a heuristic estimate to the goal. Widely used in robot path planning for grid-based maps.; Grid-Based Planning lesson
Actuator: A device that converts energy into physical motion. Common actuators in robotics include electric motors, servos, hydraulic cylinders, and pneumatic actuators. They are responsible for moving joints, wheels, and end effectors.; Sensors & Actuators lesson

B

Behavior Tree: A hierarchical model for organizing and switching between robot behaviors. Nodes include selectors (try alternatives), sequences (do in order), and conditions (check state). More modular than finite state machines for complex decision-making.; Behavior Trees lesson

C

Camera Intrinsics: Internal parameters of a camera that define how 3D points project onto the 2D image plane. Includes focal length (fx, fy), principal point (cx, cy), and distortion coefficients. Essential for converting between pixel coordinates and real-world measurements.; Camera Basics lesson
Closed-Loop Control: A control strategy that uses sensor feedback to continuously adjust commands. The controller compares the desired state (setpoint) with the measured state (feedback) and corrects the error. PID control is the most common form.; PID Control lesson
Costmap: A grid representation of the environment where each cell has a cost value indicating how expensive or dangerous it is to traverse. Used by path planners to avoid obstacles, prefer open areas, and maintain safety margins. Common types include local costmaps (from sensor data) and global costmaps (from maps).; Path Planning lesson

D

Dead Reckoning: Estimating a robot's current position by integrating its velocity and heading over time from a known starting point. Accumulates drift error over time, so it must be corrected with external references like GPS or landmark detection.; Coordinate Frames lesson
Degrees of Freedom (DoF): The number of independent parameters that define a robot's configuration. A mobile robot on a flat surface has 3 DoF (x, y, heading). A free-floating object in 3D has 6 DoF (3 translation + 3 rotation). A robot arm's DoF equals its number of independent joints.; Sensors & Actuators lesson
Depth Image: An image where each pixel stores the distance from the camera to the nearest surface rather than color. Produced by depth cameras (Intel RealSense, Kinect) or stereo vision. Used for obstacle detection, 3D reconstruction, and point cloud generation.; Depth Perception lesson
Differential Drive: A two-wheeled drive system where each wheel is independently powered. Turning is achieved by spinning the wheels at different speeds. The most common mobile robot drivetrain — simple, cheap, and effective. Includes a caster or ball for stability.; Motor Control lesson

E

EKF (Extended Kalman Filter): A nonlinear extension of the Kalman filter that linearizes the system model at each time step. Widely used for robot state estimation (fusing IMU, odometry, GPS) and in EKF-SLAM for simultaneous localization and mapping.; Localization Basics lesson
Encoder: A sensor that measures the rotation angle or position of a motor shaft. Optical encoders count light pulses through a slotted disc; magnetic encoders detect pole changes. Used for wheel odometry, joint angle feedback, and velocity measurement.; Sensors & Actuators lesson
End Effector: The device at the end of a robot arm that interacts with the environment. Common types include grippers (parallel jaw, suction), welding torches, spray nozzles, and tool changers. The choice of end effector determines what tasks the arm can perform.; Sensors & Actuators lesson

F

FIFO (First In, First Out): A queue discipline where the first message or data point enqueued is the first one processed. In robotics middleware, FIFO queues buffer sensor data between publishers and subscribers. Queue depth must be tuned to balance latency vs. data loss.
Forward Kinematics: Computing the position and orientation of a robot's end effector given its joint angles. Uses the chain of joint transforms (DH parameters or URDF link/joint tree). The inverse problem — finding joint angles for a desired end effector pose — is called inverse kinematics.; Transforms lesson
Frame (Coordinate Frame): A coordinate system attached to a specific part of the robot or environment, defined by an origin and three orthogonal axes (x, y, z). Robots use many frames: base_link, odom, map, camera_link, each related by transforms. The TF tree organizes all frame relationships.; Coordinate Frames lesson

G

Gazebo: An open-source 3D robotics simulator that provides physics simulation (ODE, Bullet, DART), sensor simulation (cameras, LiDAR, IMU), and world modeling. Used extensively for testing robot software before deploying on real hardware.
Global Planner: A path planning component that finds a route from the robot's current position to a distant goal using a map of the entire environment. Produces a path for the local planner to follow. Common algorithms: A*, Dijkstra, RRT.; Local vs Global Planning lesson
GNSS / GPS: Global Navigation Satellite System — a system of satellites that provides position, velocity, and time information to receivers on the ground. GPS (US), GLONASS (Russia), Galileo (EU), and BeiDou (China) are GNSS systems. Accuracy is typically 2-10 meters, improved by RTK to centimeter-level.

H

Homogeneous Transform: A 4x4 matrix that represents both rotation and translation in a single operation. Combines a 3x3 rotation matrix and a 3x1 translation vector. Multiplying transforms chains coordinate frame relationships, making it fundamental for robot kinematics.; Transforms lesson

I

IMU (Inertial Measurement Unit): A sensor package that measures acceleration (accelerometer), angular velocity (gyroscope), and sometimes magnetic heading (magnetometer). Provides high-frequency motion data for state estimation, orientation tracking, and fall detection. Subject to drift over time.; Sensors & Actuators lesson
Inverse Kinematics (IK): Computing the joint angles needed to place a robot's end effector at a desired position and orientation. Generally harder than forward kinematics — solutions may be non-unique, may not exist, or may require iterative numerical methods.

K

Kalman Filter: An optimal recursive estimator for linear systems that fuses noisy sensor measurements with a prediction model. Produces a statistically optimal estimate of the system state and its uncertainty. The foundation for many robotics state estimation algorithms.; Localization Basics lesson

L

LiDAR: Light Detection and Ranging — a sensor that measures distance by timing laser pulses. 2D LiDAR produces a single scan plane; 3D LiDAR produces a point cloud. Used for obstacle detection, mapping, and localization. Typical range: 10-100 meters with millimeter precision.; LiDAR & Point Clouds lesson
Local Planner: A path planning component that generates velocity commands to follow the global path while avoiding dynamic obstacles detected by sensors. Runs at high frequency (10-20 Hz). Common algorithms: DWA (Dynamic Window Approach), TEB (Timed Elastic Band).; Local vs Global Planning lesson
Localization: Determining the robot's position and orientation within a known map. Methods include particle filters (AMCL), scan matching, visual odometry, and GPS. The "L" in SLAM — one of the fundamental problems in mobile robotics.; Localization Basics lesson
Loop Closure: Detecting when a robot has returned to a previously visited location during mapping. Corrects accumulated drift by constraining the map graph. Critical for accurate SLAM — without loop closure, maps distort over long trajectories.; Loop Closure lesson

M

Message: A structured data packet sent between robot software components. Messages have defined types (e.g., Twist for velocity, LaserScan for LiDAR data, Image for camera frames) and are transmitted over topics using publish/subscribe communication.; Message Types lesson

N

Node: A single process or computation unit in a robot software system. Each node typically handles one responsibility (reading a sensor, planning a path, controlling a motor). Nodes communicate through topics, services, and actions.; Software Architecture lesson

O

Occupancy Grid: A 2D grid map where each cell stores the probability that it is occupied by an obstacle. Values range from 0 (free) to 100 (occupied), with -1 for unknown. The standard map representation for 2D navigation and SLAM.; Occupancy Grid Mapping lesson
Odometry: Estimating a robot's position change over time using wheel encoder readings or visual features. Provides relative motion (how far the robot moved) but drifts over time due to wheel slip, encoder resolution, and terrain variations.; Coordinate Frames lesson
Open-Loop Control: A control strategy that sends commands without using sensor feedback to verify the result. Simple but unreliable — the robot cannot correct for disturbances. Example: driving forward for 2 seconds and hoping you reach the wall.; PID Control lesson

P

Particle Filter: A probabilistic algorithm that represents the robot's possible positions as a set of weighted "particles" (hypotheses). Particles are resampled based on how well they match sensor observations. Used in AMCL for mobile robot localization.; Localization Basics lesson
PID Controller: Proportional-Integral-Derivative controller — the most common feedback controller in robotics. P corrects current error, I corrects accumulated error, D damps oscillation. Tuning the three gains (Kp, Ki, Kd) determines response speed, stability, and steady-state accuracy.; PID Control lesson
Point Cloud: A collection of 3D points (x, y, z) representing surfaces in the environment. Produced by 3D LiDAR, depth cameras, or stereo vision. Used for obstacle detection, 3D mapping, object recognition, and surface reconstruction.; LiDAR & Point Clouds lesson
Publish/Subscribe (Pub/Sub): A communication pattern where publishers send messages to named topics and subscribers receive messages from those topics, without either side knowing about the other. Decouples components and allows one-to-many data distribution. The primary communication pattern in ROS.; Publish/Subscribe lesson

Q

Quaternion: A four-component number (x, y, z, w) used to represent 3D rotations without gimbal lock. More compact than rotation matrices (4 values vs. 9) and smoothly interpolatable. The standard rotation representation in robotics middleware.; Quaternions lesson

R

ROS (Robot Operating System): An open-source robotics middleware framework that provides communication infrastructure (topics, services, actions), package management, hardware abstraction, and a large ecosystem of reusable packages. ROS 2 is the current generation, built on DDS for real-time and multi-robot support.
RRT (Rapidly-exploring Random Tree): A sampling-based path planning algorithm that grows a tree of random configurations toward unexplored space. Efficient for high-dimensional planning (robot arms, complex environments). RRT* is an asymptotically optimal variant.; Sampling-Based Planning lesson

S

Sensor Fusion: Combining data from multiple sensors to produce a more accurate, complete, or reliable estimate than any single sensor alone. Common examples: fusing IMU + wheel odometry for pose estimation, or fusing camera + LiDAR for object detection.; Sensor Fusion lesson
Servo Motor: A motor with built-in position feedback (encoder) and a controller that maintains a commanded angle or velocity. Common in robot arms and hobby robots. Available in standard (180 degree) and continuous rotation variants.; Sensors & Actuators lesson
Sim2Real: Transferring robot behaviors, policies, or models trained in simulation to real physical hardware. Key challenges include the reality gap (differences between simulated and real physics, sensors, and dynamics) and domain randomization to improve transfer robustness.
SLAM (Simultaneous Localization and Mapping): The problem of building a map of an unknown environment while simultaneously tracking the robot's location within it. A chicken-and-egg problem — you need a map to localize, but you need to be localized to build a map. Solved by probabilistic methods like graph-based SLAM, EKF-SLAM, and particle filter SLAM.; SLAM Explained lesson
State Machine: A model of computation where the robot is always in exactly one of a finite number of states. Transitions between states are triggered by events or conditions. Simple and predictable, but can become unwieldy for complex behaviors (state explosion problem).; State Machines lesson
Stepper Motor: A motor that moves in discrete angular increments (steps). Each electrical pulse rotates the shaft by a fixed angle (typically 1.8 degrees). Provides precise open-loop position control without an encoder. Common in CNC machines, 3D printers, and precision robotics.

T

TF Tree (Transform Tree): A hierarchical data structure that maintains the spatial relationships between all coordinate frames on a robot. Allows querying the transform between any two frames at any point in time. Essential for converting sensor data between different reference frames.; Transform Trees lesson
Topic: A named communication channel in publish/subscribe systems. Publishers send messages to a topic; all subscribers to that topic receive a copy. Topics are typed (each carries one message type) and named by convention (e.g., /cmd_vel for velocity commands, /scan for LiDAR data).; Publish/Subscribe lesson
Twist: A message type that represents velocity in free space, with linear velocity (x, y, z in m/s) and angular velocity (x, y, z in rad/s). The standard way to command mobile robot motion — published on /cmd_vel.; Velocity Commands lesson

U

URDF (Unified Robot Description Format): An XML format for describing a robot's physical structure — links (rigid bodies with visual/collision geometry and inertia), joints (revolute, prismatic, fixed, continuous), and their hierarchical relationships. Used by simulators, visualization tools, and motion planners.; Packaging lesson

V

Velocity Command: An instruction sent to the robot's motor controller specifying desired linear and angular velocities. Typically published as a Twist message on the /cmd_vel topic. The navigation stack produces velocity commands; the motor driver executes them.; Velocity Commands lesson
Visual Odometry: Estimating a robot's motion by tracking visual features (corners, edges) across sequential camera images. More robust than wheel odometry on slippery surfaces, but computationally expensive and affected by lighting changes and featureless environments.
Voxel: A 3D volume element — the 3D analog of a pixel. Voxel grids discretize 3D space into cubes, each storing occupancy or other properties. Used for 3D costmaps, OctoMap representations, and downsampling point clouds.

W

Waypoint: A specific position (and optionally orientation) that a robot should pass through on its way to a goal. A path is often represented as a sequence of waypoints. The path planner generates waypoints; the local planner navigates between them.; Local vs Global Planning lesson

X

XACRO: An XML macro language that extends URDF with variables, math expressions, and reusable macros. Reduces repetition and makes complex robot descriptions manageable. XACRO files are preprocessed into plain URDF before use.; Packaging lesson

Y

Yaw / Pitch / Roll: Euler angle convention for describing 3D orientation. Yaw rotates around the vertical axis (heading), pitch tilts forward/backward, roll tilts left/right. Intuitive but suffers from gimbal lock — quaternions are preferred for computation.; Quaternions lesson