SDF (Simulation Description Format)

SDF is the de facto standard for describing experimental robotic platforms in research labs worldwide. Its support for nested models, complex kinematic trees, and detailed inertial properties allows researchers to accurately replicate real robots like Boston Dynamics' Spot or Universal Robots' arms in simulation. This enables:

• Reinforcement Learning (RL): Training policies in simulators like Gazebo or Ignition before real-world deployment.
• Algorithm Benchmarking: Providing a consistent, reproducible description of robots and test worlds for comparing motion planning or control algorithms.
• Prototyping New Designs: Rapidly iterating on virtual robot concepts by editing SDF files, without physical fabrication costs.

SDF is used to construct vast, detailed virtual proving grounds for self-driving cars and drones. It can define:

• High-Fidelity Urban Worlds: Complete with roads, traffic signs, buildings, and dynamic pedestrians (also defined as SDF models).
• Sensor Suites: Precisely modeling the intrinsic and extrinsic parameters of LiDAR, cameras, radar, and IMUs mounted on the vehicle.
• Environmental Conditions: Configuring atmospheric effects, time-of-day lighting, and weather phenomena (e.g., fog, rain) to test perception stack robustness. Platforms like NVIDIA Isaac Sim and CARLA (which uses SDF-like descriptions) leverage this capability to run millions of miles of driving simulation, testing edge cases far too dangerous for physical testing.

In manufacturing and logistics, SDF is used to build digital twins of entire workcells or warehouses. It describes:

• Production Lines: Conveyors, robotic arms (e.g., defined via URDF and integrated into SDF), fixtures, and the parts being assembled.
• Physics of Interaction: Defining contact materials with specific friction and restitution coefficients to simulate part grasping, dropping, or sliding accurately.
• Logical Sensors: Modeling proximity sensors, photoelectric gates, and force-torque sensors that trigger automated processes. This allows engineers to validate cycle times, prevent collisions, and optimize layouts in simulation before deploying or modifying expensive physical infrastructure.

SDF is critical for simulating robots destined for environments where physical testing is impossible or prohibitively expensive, such as space, deep sea, or nuclear facilities.

• Planetary Rovers: Models for NASA's Robonaut or ESA's rover prototypes include detailed wheel-terrain interaction models and specialized manipulators in their SDF descriptions.
• Underwater ROVs: Simulating hydrodynamic drag and buoyancy forces by applying custom plugins to SDF models.
• Degraded Communication: Testing autonomy stacks under simulated latency and dropouts characteristic of deep-space operations. The format's ability to integrate custom physics plugins via its <plugin> tag is essential for modeling non-terrestrial physics.

SDF is used to create simulated environments where humans and robots co-exist, enabling safe testing of collaborative robots (cobots).

• Human Model Integration: While often simplified, human avatars can be defined in SDF to test proximity awareness and collision avoidance algorithms.
• Workspace Simulation: Defining shared workspaces, tools, and task objects to validate that a robot's planned trajectory does not intrude into a human's safe zone.
• Sensor Simulation for HRI: Modeling depth cameras and microphone arrays used for human pose estimation and voice command recognition in context. This simulation-first approach is mandated by emerging safety standards (e.g., ISO/TS 15066) for cobot deployment.

Beyond direct use in simulators, SDF serves as the interlingua for a vast ecosystem of robotics development tools.

• Model Repositories: Centralized libraries like the Gazebo Model Database host thousands of pre-built SDF models of common robots, objects, and environments.
• Conversion Tools: URDF to SDF converters are essential pipelines, as many robots are first described in URDF for basic visualization, then converted to SDF for high-fidelity simulation with proper physics.
• Configuration Management: SDF files are version-controlled assets in a robotics project, defining the exact simulation configuration used for a particular test or training run, ensuring full reproducibility. This tooling ecosystem solidifies SDF's role as the lingua franca for describing simulated worlds.

URDF is an XML-based file format used to describe the kinematic tree, visual geometry, and inertial properties of a robot. It is the predecessor and common companion to SDF.

• Core Purpose: Primarily defines a robot's kinematic structure (links and joints) for visualization and basic simulation in tools like RViz and Gazebo.
• Key Limitation: Designed as a single-robot description. It cannot natively describe complete worlds with multiple robots, lighting, or terrain.
• Relationship to SDF: SDF is a superset format. Many simulation workflows use a robot's URDF, which is then automatically converted or wrapped into an SDF model for use in a full world simulation. SDF supports URDF features and adds world-level description capabilities.

MuJoCo (Multi-Joint dynamics with Contact) is a high-performance physics engine. Its native MJCF (MuJoCo Modeling Format) is a direct XML-based competitor and alternative to SDF for describing robots and scenes.

• Comparison to SDF: Both are hierarchical, XML-based scene description formats. MJCF is tightly integrated with and optimized for the MuJoCo engine's constraint-based solver.
• Key Differences: MJCF often uses more concise syntax for defining geometries and inertia. It has first-class support for tendons and equality constraints. SDF has broader adoption across multiple simulators (Gazebo, Ignition) and stronger support for world-building (lighting, atmospheric effects).
• Conversion: Tools exist to convert between URDF/SDF and MJCF, though fidelity may be lost due to differing solver models.

Gazebo is the flagship open-source robotics simulator where SDF was originally developed and remains the primary world description format. The simulator runtime parses SDF files to instantiate the simulation.

• Native Format: SDF is Gazebo's native scene description language. The simulator provides the GUI tools, plugin system, and physics backend that bring an SDF world to life.
• Evolution: The format evolved from SDF 1.0 (used in "Gazebo classic") to SDF 1.7+ in the newer Ignition Gazebo (now renamed back to Gazebo). The newer versions support more modular model composition, improved performance, and better sensor modeling.
• Ecosystem: Gazebo demonstrates SDF's full capabilities, including nested models, scripted scenarios, and plugin-based custom logic for sensors and actuators.