Skip to main content

Glossary of Terms

Quick reference for technical terms used throughout the Physical AI Humanoid Robotics course.

A​

Action (ROS 2) : Long-running tasks in ROS 2 with feedback and cancellation. Unlike services, actions provide intermediate feedback during execution.

Actuator : A mechanical component that moves or controls a mechanism (motors, servos, hydraulics).

ADAS (Advanced Driver Assistance Systems) : Vehicle systems that use sensors and AI for safety features like lane keeping and collision avoidance.

B​

Bipedal Locomotion : Walking or movement using two legs, a key challenge in humanoid robotics requiring balance and stability control.

BNO055 : A 9-axis Inertial Measurement Unit (IMU) combining accelerometer, gyroscope, and magnetometer.

C​

CapEx (Capital Expenditure) : Upfront costs for purchasing equipment (workstations, robots) vs. OpEx (operational costs).

Colcon : The build system for ROS 2 packages, replacing catkin from ROS 1.

Computer Vision : Field of AI enabling computers to interpret and understand visual information from images and videos.

CUDA (Compute Unified Device Architecture) : NVIDIA's parallel computing platform for GPU-accelerated applications.

D​

Depth Camera : Camera that measures distance to objects (e.g., Intel RealSense D435i) using stereo vision or structured light.

Digital Twin : A virtual replica of a physical robot used for simulation, testing, and training before real-world deployment.

DOF (Degrees of Freedom) : Number of independent movements a robot joint or system can perform. Humanoids typically have 20-40+ DOF.

E​

Embodied AI : Artificial intelligence systems deployed in physical robots that interact with the real world through sensors and actuators.

End Effector : The device at the end of a robotic arm (gripper, hand, tool) that interacts with objects.

F​

Forward Kinematics : Calculating the position and orientation of a robot's end effector given joint angles.

G​

Gazebo : Open-source physics-based robot simulator widely used with ROS for testing before hardware deployment.

Grasping : Robot capability to pick up and manipulate objects using grippers or hands.

GPT (Generative Pre-trained Transformer) : Large language models (like GPT-4) used for natural language understanding in conversational robotics.

H​

HRI (Human-Robot Interaction) : Field studying how humans and robots communicate and collaborate effectively.

Humanoid Robot : Robot with a human-like body structure (head, torso, two arms, two legs).

I​

IMU (Inertial Measurement Unit) : Sensor measuring acceleration, rotation, and magnetic field - critical for robot balance and localization.

Inference : Running a trained AI model to make predictions (vs. training the model).

Inverse Kinematics : Calculating joint angles needed to position a robot's end effector at a desired location.

Isaac ROS : NVIDIA's hardware-accelerated ROS 2 packages for perception, navigation, and manipulation.

Isaac Sim : NVIDIA's photorealistic robot simulator built on Omniverse, optimized for AI training.

J​

Jetson : NVIDIA's embedded computing platform for edge AI (Jetson Orin Nano, Orin NX, etc.).

Joint : Connection point between robot links allowing movement (revolute, prismatic, spherical).

K​

Kinematics : Study of robot motion without considering forces (positions, velocities, accelerations).

L​

Launch File : ROS 2 file (Python or XML) that starts multiple nodes with parameters and configurations.

Lidar (Light Detection and Ranging) : Sensor using laser pulses to measure distances and create 3D maps of environments.

Locomotion : Robot movement capability (walking, running, jumping for humanoids).

M​

Manipulation : Robot's ability to interact with and move objects using arms and grippers.

MoveIt : ROS framework for motion planning, manipulation, and collision avoidance (MoveIt 2 for ROS 2).

N​

Nav2 : ROS 2 navigation stack for autonomous mobile robot navigation with obstacle avoidance.

Node (ROS 2) : Independent process that performs computation and communicates via topics, services, or actions.

O​

Omniverse : NVIDIA's platform for 3D simulation and collaboration, foundation for Isaac Sim.

ONNX (Open Neural Network Exchange) : Standard format for representing machine learning models across frameworks.

OpEx (Operational Expenditure) : Recurring costs (cloud instances, maintenance) vs. CapEx (upfront equipment purchases).

P​

Package (ROS 2) : Organizational unit containing nodes, libraries, configuration files, and launch files.

Perception : Robot's ability to sense and interpret its environment (vision, LIDAR, depth sensing).

Physical AI : AI systems deployed in physical embodiments (robots) that interact with the real world.

Publish-Subscribe : ROS 2 communication pattern where nodes publish data to topics that other nodes subscribe to.

Q​

QoS (Quality of Service) : ROS 2 policies controlling message reliability, durability, and delivery guarantees.

R​

RealSense : Intel's line of depth cameras (D435i, D455) with RGB, depth, and IMU sensors.

Reinforcement Learning (RL) : Machine learning approach where agents learn optimal behaviors through trial and error.

RGB-D : Image data combining RGB color with Depth information.

Robot Description : Formal specification of robot geometry, links, joints, and sensors (URDF, SDF, USD).

ROS (Robot Operating System) : Middleware framework for robot software development. ROS 2 is the latest version.

rviz2 : ROS 2 visualization tool for displaying sensor data, robot models, and trajectories.

S​

SDF (Simulation Description Format) : XML format for describing simulation worlds and robots in Gazebo.

Sensor Fusion : Combining data from multiple sensors (camera, LIDAR, IMU) for more accurate perception.

Service (ROS 2) : Request-response communication pattern for quick, synchronous operations.

Sim-to-Real : Process of training robots in simulation and transferring learned behaviors to physical hardware.

SLAM (Simultaneous Localization and Mapping) : Robot capability to build maps while tracking its own position.

T​

TensorRT : NVIDIA's SDK for high-performance deep learning inference optimization.

Topic (ROS 2) : Named channel for asynchronous data streaming between nodes.

TOPS (Tera Operations Per Second) : Measure of AI processing performance (Jetson Orin Nano: 40 TOPS).

U​

Unity : Game engine used for high-fidelity robot simulation with Unity Robotics Hub.

URDF (Unified Robot Description Format) : XML format for describing robot kinematics, dynamics, and visualization in ROS.

USD (Universal Scene Description) : Pixar's format for 3D scenes, used in Isaac Sim for robot and environment representation.

V​

VLA (Vision-Language-Action) : AI models that combine visual perception, natural language understanding, and physical action planning.

VSLAM (Visual SLAM) : SLAM using camera images (vs. LIDAR SLAM).

W​

Whisper : OpenAI's speech recognition model for converting voice to text.

Workspace (ROS 2) : Directory structure containing ROS 2 packages, build files, and installation.

X​

Xacro : XML macro language for simplifying URDF files with variables and reusable components.

Y​

YOLO (You Only Look Once) : Real-time object detection algorithm family (YOLOv5, YOLOv8) for computer vision.

Z​

ZMP (Zero Moment Point) : Concept in bipedal robotics for maintaining balance by controlling where weight is distributed.