Glossary
A comprehensive list of technical terms used throughout this textbook.
A
Action: In ROS 2, a goal-based asynchronous communication pattern with feedback and cancellation support.
Action Tokenization: Representing continuous robot actions as discrete tokens for transformer-based VLA models.
ArticulationBody: Unity component for simulating robot joints with physics.
B
Bounding Box: Rectangle enclosing a detected object in image space.
C
Colcon: Build tool for ROS 2 packages and workspaces.
Costmap: 2D occupancy grid representing traversability for mobile robot navigation.
CUDA: NVIDIA's parallel computing platform and API for GPU programming.
cuMotion: NVIDIA's GPU-accelerated motion planning library.
D
DART: Physics engine used in Gazebo, well-suited for manipulation tasks.
DDS: Data Distribution Service, the middleware protocol underlying ROS 2 communication.
Depth Camera: Sensor that captures both RGB images and per-pixel depth information.
Digital Twin: High-fidelity virtual replica of a physical robot system.
DNN Inference: Running a trained deep neural network on new input data.
Domain Randomization: Varying simulation parameters (lighting, textures, physics) to improve sim-to-real transfer.
E
Embodied Intelligence: AI systems that interact with and learn from the physical world.
F
Foundation Model: Large pre-trained model (like GPT-4) transferable to many downstream tasks.
Function Calling: Structured LLM outputs formatted as function/method calls with parameters.
G
Gazebo: Open-source robot simulator with physics engines and sensor models.
I
IMU: Inertial Measurement Unit, sensor measuring acceleration and angular velocity.
INT8 Quantization: Reducing neural network weights from 32-bit to 8-bit integers for faster inference.
Isaac Gym: NVIDIA's GPU-accelerated physics simulator for massively parallel reinforcement learning.
Isaac ROS: Collection of GPU-accelerated ROS 2 packages for robotics perception and planning.
Isaac Sim: NVIDIA's high-fidelity robot simulator based on Omniverse.
Isaac SDK: NVIDIA's development framework for AI-powered robotics applications.
L
Launch File: Script that starts multiple ROS 2 nodes with specified configuration.
LiDAR: Light Detection and Ranging sensor that generates 3D point clouds.
LLM: Large Language Model (e.g., GPT-4, Claude) for natural language understanding and generation.
Loop Closure: In SLAM, detecting when the robot revisits a previously mapped location.
M
Multimodal: Systems processing multiple input types (vision, language, touch, etc.) simultaneously.
N
Namespace: Hierarchical grouping mechanism for ROS 2 nodes and topics.
Nav2: ROS 2 navigation framework for autonomous mobile robots.
Node: In ROS 2, an independent process performing computation.
Node Composition: Running multiple ROS 2 nodes in a single process for efficiency.
O
ONNX: Open Neural Network Exchange, a format for representing deep learning models.
P
Package: In ROS 2, a collection of nodes, libraries, configuration files, and dependencies.
package.xml: Manifest file declaring ROS 2 package metadata and dependencies.
Parameter: Runtime configuration value for ROS 2 nodes.
Physics Engine: Software simulating forces, collisions, and dynamics in virtual environments.
Point Cloud: 3D representation of space as a set of points with (x, y, z) coordinates.
PPO: Proximal Policy Optimization, a popular reinforcement learning algorithm.
Prompt Engineering: Designing input text to guide LLM behavior and outputs.
Publisher: ROS 2 node component that sends messages to a topic.
Q
QoS: Quality of Service policies controlling message delivery reliability and history in ROS 2.
R
ROS-TCP-Connector: Unity package enabling communication between Unity and ROS 2.
ros_gz_bridge: ROS 2 package for bidirectional communication with Gazebo.
S
SDF: Simulation Description Format used by Gazebo to define worlds and models.
Service: In ROS 2, synchronous request-response communication pattern.
Sim-to-Real Gap: Differences between simulated and real-world robot behavior.
Sim-to-Real Transfer: Techniques for deploying policies trained in simulation to physical robots.
SLAM: Simultaneous Localization and Mapping, building a map while determining robot position.
STT: Speech-to-Text conversion.
Subscriber: ROS 2 node component that receives messages from a topic.
T
TensorRT: NVIDIA's library for optimizing and deploying deep neural networks.
tf2: ROS 2 library for tracking coordinate frame transformations over time.
Topic: In ROS 2, a named channel for asynchronous message passing.
U
Unity: Game engine adapted for robotics simulation and visualization.
URDF: Unified Robot Description Format, XML specification for robot kinematics and geometry.
V
VAD: Voice Activity Detection, identifying when speech is occurring in audio.
Visual SLAM (VSLAM): SLAM using camera images as the primary sensor.
VLA: Vision-Language-Action model, end-to-end transformer for robot control.
W
Wake Word: Trigger phrase that activates voice control systems.
Whisper: OpenAI's speech-to-text model for voice transcription.
Workspace: Directory containing one or more ROS 2 packages.
Total Terms: 48+
For term usage in context, see the chapter where each term is introduced (linked in the Key Terms sections).