Robotics Glossary
Quick reference for robotics terminology used throughout the Physical AI & Humanoid Robotics course. Terms are organized alphabetically.
A
Action (ROS 2)
Asynchronous communication pattern in ROS 2 for long-running tasks (e.g., navigation to a goal). Actions provide feedback during execution and can be preempted. Consists of goal, result, and feedback messages.
Related: Module 1: ROS 2
Actuator
Device that converts energy (electrical, pneumatic, hydraulic) into motion. Examples: electric motors, servo motors, pneumatic cylinders. Actuators enable robots to move and manipulate objects.
Algolia DocSearch
Search-as-a-service solution for documentation websites. Provides instant search with typo tolerance, faceting, and custom ranking. Used in this textbook for content search.
Related: Introduction
B
Bag File (ROS 2)
Format for recording and playing back ROS 2 message data. Essential for debugging, algorithm development, and sharing datasets. File extension: .db3 (SQLite format in ROS 2).
Related: Module 1: ROS 2
Base Link
Standard coordinate frame attached to a robot's base (typically center of robot at ground level). All other robot frames are defined relative to base_link.
Related: Module 1: ROS 2
C
Capstone Project
Culminating project integrating all course modules. Students build an autonomous humanoid system with voice-driven manipulation: Voice → Plan → Navigate → Perceive → Manipulate.
Related: Module 4: VLA & Humanoids
Computation Graph
Network of ROS 2 nodes communicating via topics, services, and actions. Visualized using rqt_graph. Enables distributed robotics architectures.
Related: Module 1: ROS 2
CUDA (Compute Unified Device Architecture)
NVIDIA parallel computing platform enabling GPU-accelerated computation. Required for Isaac Sim and deep learning inference on robots.
Related: Workstation Setup
D
DDS (Data Distribution Service)
Middleware standard used by ROS 2 for real-time, reliable communication. Implementations include Fast-DDS, CycloneDDS, RTI Connext. Replaces TCPROS from ROS 1.
Related: Module 1: ROS 2
Degrees of Freedom (DOF)
Number of independent parameters defining robot configuration. A humanoid arm typically has 7 DOF (shoulder: 3, elbow: 1, wrist: 3).
Related: Module 4: VLA & Humanoids
Digital Twin
Virtual replica of a physical system used for simulation, testing, and optimization. In robotics, digital twins enable algorithm validation before hardware deployment.
Related: Module 2: Digital Twin
Docusaurus
Static site generator used to build this textbook. Features include React-based theming, Algolia search, and MDX support.
E
Embodied AI
AI systems that interact with the physical world through sensors and actuators. Combines perception, reasoning, and action execution. Contrast with purely digital AI (e.g., chatbots).
Related: Module 4: VLA & Humanoids
End Effector
Terminal device attached to robot's kinematic chain (e.g., gripper, suction cup, camera). Performs manipulation tasks.
Related: Module 4: VLA & Humanoids
F
Flexsearch
Fast, memory-efficient full-text search library. Used in this textbook for glossary search. Provides sub-100ms query times.
Related: This glossary uses Flexsearch for instant term lookup
Forward Kinematics (FK)
Computing end-effector pose given joint angles. Used for simulation and motion visualization.
Related: Module 4: VLA & Humanoids
G
Gazebo
Open-source 3D robotics simulator with physics engines (ODE, Bullet, Dart). Integrates natively with ROS 2. Suitable for rapid prototyping.
Related: Module 2: Digital Twin
GPU (Graphics Processing Unit)
Parallel processor designed for graphics rendering. Repurposed for AI training (CUDA), physics simulation (PhysX), and ray tracing. NVIDIA RTX GPUs required for Isaac Sim.
Related: Module 3: Isaac Sim
H
Homogeneous Transformation
4x4 matrix combining rotation (3x3) and translation (3x1) for coordinate frame transformations. Standard in robotics for pose representation.
Related: Module 4: VLA & Humanoids
Humanoid Robot
Robot with human-like morphology (head, torso, two arms, two legs). Examples: Boston Dynamics Atlas, Unitree H1, Tesla Optimus. Typically 30+ degrees of freedom.
Related: Module 4: VLA & Humanoids
I
IMU (Inertial Measurement Unit)
Sensor measuring acceleration and angular velocity. Combines accelerometer, gyroscope, and sometimes magnetometer. Essential for robot balance and orientation estimation.
Related: Module 2: Digital Twin
Inverse Kinematics (IK)
Computing joint angles needed to achieve desired end-effector pose. More complex than FK; may have multiple solutions or no solution.
Related: Module 4: VLA & Humanoids
Isaac Gym
NVIDIA's reinforcement learning toolkit built on Isaac Sim. Enables massively parallel policy training (10,000+ environments simultaneously) using GPU acceleration.
Related: Module 3: Isaac Sim
Isaac Sim
NVIDIA's GPU-accelerated robotics simulator built on Omniverse. Features photorealistic RTX rendering, PhysX 5 physics, native ROS 2 integration.
Related: Module 3: Isaac Sim
J
Jacobian Matrix
Matrix relating joint velocities to end-effector velocities. Used for velocity control and singularity analysis. Size: 6×n (6 DOF end-effector, n joints).
Related: Module 4: VLA & Humanoids
Jetson Orin Nano
NVIDIA embedded GPU platform for edge AI. Features 1024 CUDA cores, 8GB RAM, 5-15W power consumption. Ideal for on-robot AI inference.
Related: Edge Kit Setup
L
LiDAR (Light Detection and Ranging)
Sensor using laser pulses to measure distances and create 3D point clouds. Used for mapping, obstacle avoidance, and localization.
Related: Module 2: Digital Twin
LLM (Large Language Model)
AI model trained on text data for natural language understanding. Examples: GPT-4, Claude, LLaMA. In robotics, used for task planning from voice commands.
Related: Module 4: VLA & Humanoids
M
Manipulation
Process of grasping, moving, and releasing objects. Requires perception (object detection), planning (grasp pose), and control (trajectory execution).
Related: Module 4: VLA & Humanoids
MoveIt 2
Motion planning framework for ROS 2. Includes inverse kinematics solvers, collision checking, and trajectory generation. Widely used for manipulation.
Related: Module 4: VLA & Humanoids
N
Nav2 (Navigation 2)
ROS 2 navigation stack for mobile robots. Provides path planning, obstacle avoidance, and recovery behaviors. Uses costmaps for environment representation.
Related: Module 3: Isaac Sim
Node (ROS 2)
Independent process performing computation (perception, planning, control). Nodes communicate via topics/services/actions. Example: camera driver node publishes to /camera/image_raw.
Related: Module 1: ROS 2
O
Occupancy Grid
2D/3D representation of environment where each cell indicates free, occupied, or unknown space. Used by Nav2 for path planning.
Related: Module 3: Isaac Sim
Odometry
Estimation of robot position based on motion sensors (wheel encoders, IMU, visual). Accumulates error over time; corrected by SLAM or GPS.
Related: Module 3: Isaac Sim
Omniverse
NVIDIA platform for 3D collaboration and simulation. Isaac Sim is built on Omniverse. Supports USD format and RTX rendering.
Related: Module 3: Isaac Sim
P
PhysX
NVIDIA's GPU-accelerated physics engine. Supports rigid bodies, soft bodies, fluids, and cloth. Used in Isaac Sim for realistic simulation.
Related: Module 3: Isaac Sim
Point Cloud
Set of 3D points representing object surfaces. Generated by LiDAR, depth cameras, or stereo vision. Format: PCL, Open3D.
Related: Module 3: Isaac Sim
Q
QoS (Quality of Service)
Configuration for ROS 2 communication reliability and performance. Profiles include: Best Effort, Reliable, Sensor Data, Services. Tunable parameters: history depth, durability, liveliness.
Related: Module 1: ROS 2
R
rclpy
ROS 2 Python client library. Used for writing nodes, publishers, subscribers, services in Python. Alternative: rclcpp (C++).
Related: Module 1: ROS 2
Reinforcement Learning (RL)
Machine learning paradigm where agents learn by trial and error through rewards. In robotics, used for grasping, manipulation, and navigation.
Related: Module 3: Isaac Sim
RGB-D Camera
Camera providing both color (RGB) and depth (D) information. Examples: Intel RealSense, Microsoft Kinect. Used for 3D perception.
Related: Module 3: Isaac Sim
ROS 2 (Robot Operating System 2)
Middleware framework for robot software development. Provides communication (DDS), tools (rviz, rqt), and libraries (tf2, nav2).
Related: Module 1: ROS 2
RTX (Ray Tracing Texel eXtreme)
NVIDIA GPU architecture with dedicated ray tracing cores. Enables photorealistic rendering in Isaac Sim for sensor simulation.
Related: Module 3: Isaac Sim
RViz2
3D visualization tool for ROS 2. Displays robot models, sensor data, transforms, and planning results. Essential debugging tool.
Related: Module 1: ROS 2
S
Service (ROS 2)
Synchronous request-response communication pattern. Client sends request, waits for server response. Used for queries and short-lived tasks.
Related: Module 1: ROS 2
SLAM (Simultaneous Localization and Mapping)
Problem of building a map while simultaneously localizing within it. Variants: LiDAR SLAM, Visual SLAM (VSLAM), RGB-D SLAM.
Related: Module 3: Isaac Sim
T
TF2 (Transform Library 2)
ROS 2 library for managing coordinate frame relationships. Tracks transforms over time, enabling conversion between frames (e.g., base_link → camera_link).
Related: Module 1: ROS 2
Topic (ROS 2)
Named communication channel for asynchronous message passing. Publishers send messages, subscribers receive. Many-to-many pattern. Example: /camera/image_raw.
Related: Module 1: ROS 2
U
Unity
Game engine used for photorealistic robot simulation. Unity Robotics Hub provides ROS integration. Alternative to Gazebo for synthetic data generation.
Related: Module 2: Digital Twin
URDF (Unified Robot Description Format)
XML format for describing robot kinematics, dynamics, and visual properties. Parsed by ROS 2 for robot_state_publisher and visualization.
Related: Module 1: ROS 2
V
VLA (Vision-Language-Action)
AI model architecture that integrates visual perception, language understanding, and physical action. Enables robots to respond to natural language commands.
Related: Module 4: VLA & Humanoids
VSLAM (Visual SLAM)
SLAM variant using camera images for localization and mapping. Algorithms: ORB-SLAM3, RTAB-Map, OpenVSLAM. Less affected by transparent/reflective surfaces than LiDAR.
Related: Module 3: Isaac Sim
X
Xacro (XML Macros)
Extension to URDF enabling parameterization and code reuse via macros. Simplifies complex robot descriptions. File extension: .xacro.
Related: Module 1: ROS 2
Z
Zero-Shot Generalization
AI model's ability to perform tasks without specific training. VLA models like RT-2 can execute novel manipulation tasks by leveraging web-scale pre-training.
Related: Module 4: VLA & Humanoids
Contributing
Found a missing term? Submit a pull request to add robotics terminology to this glossary.
Search Tips:
- Use the search box above to find terms quickly
- Terms link to relevant course modules for deeper learning
- Alphabetically organized for easy browsing