Skip to main content

Robotics Terminology Glossary

This glossary defines key terms and concepts used throughout the Physical AI & Humanoid Robotics course, including NVIDIA Isaac, Digital Twins, and AI-driven robotic systems.

Core Digital Twin Concepts​

Digital Twin​

A virtual representation of a physical robot system that enables safe testing of perception, interaction, and motion. It continuously synchronizes with the physical system through real-time data and uses simulation, machine learning, and reasoning to help make decisions.

Physics Simulation​

Computational modeling of physical laws (gravity, collisions, dynamics) to create realistic robot behavior in virtual environments. This includes simulating forces, motion, and interactions between objects.

Visualization Engine​

Software component responsible for rendering visual elements and enabling human-robot interaction in simulation environments. Examples include Unity for visual realism and Gazebo for basic visualization.

Sensor Simulation​

Virtual modeling of real-world sensors to test perception and state estimation algorithms in digital environments. This includes simulating LiDAR, cameras, IMUs, and other robotic sensors.

Gazebo-Specific Terms​

SDF (Simulation Description Format)​

XML-based format used by Gazebo to describe simulation worlds, models, and physics properties. It defines objects, environments, and their physical characteristics.

URDF (Unified Robot Description Format)​

XML format used to describe robot models, including links, joints, and kinematic chains. Gazebo integrates with URDF to simulate robot models.

Physics Engine​

Computational system that simulates physical laws such as gravity, collisions, and dynamics. Gazebo supports multiple engines including ODE, Bullet, and Simbody.

Sensor Plugin​

Custom code that extends Gazebo's capabilities to simulate specific types of sensors, processing their data and publishing it to ROS/ROS2 topics.

Collision Mesh​

Geometric representation of an object used specifically for collision detection, often simplified compared to the visual mesh for performance reasons.

Unity-Specific Terms​

Real-time Rendering​

Process of generating computer graphics at speeds fast enough to display them as continuous motion, typically at 30-60 frames per second for smooth interaction.

Physically Based Rendering (PBR)​

Rendering approach that simulates light interaction with surfaces using physical principles to achieve realistic material appearance.

Asset Pipeline​

Workflow and tools used to import, process, and optimize 3D models, textures, and other content for use in Unity applications.

LOD (Level of Detail)​

Technique that uses different versions of a 3D model with varying complexity based on distance from the viewer to optimize performance.

Unity Robotics Hub​

Collection of tools, samples, and documentation that facilitate robotics development in Unity, including ROS# communication bridge.

Sensor Simulation Terms​

Noise Model​

Mathematical representation of the random variations in sensor measurements that occur in real-world conditions, including Gaussian noise, quantization, and bias.

Ground Truth​

Reference data representing the actual state of the simulated environment, used for validating sensor accuracy and algorithm performance.

Sensor Fusion​

Process of combining data from multiple sensors to achieve better accuracy and reliability than individual sensors could provide.

Field of View (FOV)​

Angular extent of the observable world that a sensor can observe at any given moment, typically measured in degrees.

Update Rate​

Frequency at which a sensor produces new measurements, typically measured in hertz (Hz).

Dynamic Range​

Ratio between the largest and smallest measurable values that a sensor can accurately detect.

Robotics and Simulation Terms​

Forward Kinematics​

Calculation of the position and orientation of a robot's end effector based on the known joint angles and link geometry.

Inverse Kinematics​

Calculation of the required joint angles to achieve a desired position and orientation of a robot's end effector.

Real-time Factor (RTF)​

Ratio of simulation time to wall-clock time, indicating how fast or slow the simulation runs relative to real time.

Domain Randomization​

Technique of randomizing simulation parameters to make models trained in simulation more robust when transferred to real-world applications.

Perception Pipeline​

Series of processing steps that transform raw sensor data into meaningful information about the environment.

NVIDIA Isaac Ecosystem Terms​

NVIDIA Isaac​

NVIDIA's comprehensive robotics platform that combines hardware, software, and simulation tools to accelerate the development and deployment of AI-powered robots. It provides hardware-accelerated solutions for robotic perception, navigation, and manipulation leveraging GPU computing and deep learning.

Isaac Sim​

NVIDIA's photorealistic simulation environment built on the Omniverse platform. It provides physics-accurate environments where robots can be designed, tested, and trained using synthetic data generation. Isaac Sim enables extensive testing of robotic algorithms before deployment on physical hardware.

Isaac ROS​

NVIDIA's collection of hardware-accelerated perception and navigation packages for the Robot Operating System (ROS). It bridges high-performance GPU computing with traditional robotics frameworks, enabling real-time processing of complex sensor data for robotic perception and navigation tasks.

Synthetic Data​

Artificially generated data that mimics real-world data characteristics without requiring physical data collection. In robotics, synthetic data is generated through simulation environments like Isaac Sim to train perception algorithms and reduce the need for extensive real-world data collection.

Hardware Acceleration (Robotics)​

The use of specialized hardware (particularly GPUs) to speed up computationally intensive algorithms in robotics, such as computer vision, sensor processing, and machine learning inference. This enables real-time performance for perception and navigation tasks.

VSLAM (Visual Simultaneous Localization and Mapping)​

A technique that uses visual sensors (cameras) to simultaneously construct a map of an unknown environment and localize the robot within that map. Hardware acceleration significantly improves VSLAM performance for real-time applications.

Isaac Apps​

Pre-built applications and reference architectures provided by NVIDIA Isaac that demonstrate best practices for implementing common robotics capabilities, such as navigation, manipulation, and perception.

Omniverse Platform​

NVIDIA's simulation and collaboration platform that powers Isaac Sim. It provides real-time physically accurate simulation, visualization, and synthetic data generation capabilities for robotics development.