In robotics, simulation is not just a visualization tool—it is where control logic, perception pipelines, and autonomy stacks are validated long before touching real hardware. Two tools form the backbone of this workflow in ROS-based systems: URDF and Gazebo.

Together, they allow engineers to define what a robot is and how it behaves in the physical world.

This post explains how URDF and Gazebo work together, why they matter in real systems, and how they are used in practical robotics development.

What is URDF?

URDF (Unified Robot Description Format) is an XML-based format used in ROS to describe a robot’s structure and kinematics.

URDF defines:

  • Rigid bodies (links)
  • Motion constraints (joints)
  • Coordinate frames
  • Visual and collision geometry
  • Mass and inertia properties

At its core, URDF answers one question:

“What is the robot made of, and how are the parts connected?”

Core URDF Building Blocks

A link represents a rigid body such as:

  • Chassis
  • Wheels
  • Sensor mounts
  • LiDAR housing
  • Camera frame

Each link can contain:

  • Visual geometry – meshes used for rendering
  • Collision geometry – simplified shapes for physics
  • Inertial properties – mass and inertia tensor
<link name="base_link">
  <visual>
    <geometry>
      <mesh filename="base.dae"/>
    </geometry>
  </visual>
  <collision>
    <geometry>
      <box size="0.6 0.4 0.2"/>
    </geometry>
  </collision>
  <inertial>
    <mass value="20"/>
    <inertia ixx="1.2" iyy="1.5" izz="2.0"/>
  </inertial>
</link>

Joints

Joints define how links move relative to each other.

Common joint types:

  • fixed – no motion (sensor mounts)
  • revolute – rotation with limits (steering)
  • continuous – infinite rotation (wheels)
  • prismatic – linear motion (actuators)
<joint name="wheel_joint" type="continuous">
  <parent link="base_link"/>
  <child link="wheel_link"/>
  <axis xyz="0 1 0"/>
</joint>

Why URDF Alone Is Not Enough

URDF describes kinematics, but it does not define:

  • Contact dynamics
  • Friction
  • Gravity response
  • Sensor noise
  • Actuator behavior

This is where Gazebo comes in.

Gazebo: Bringing Physics to Life

Gazebo is a 3D physics simulator that integrates tightly with ROS. It takes a URDF model and adds real-world physics on top of it.

Gazebo simulates:

  • Gravity and collisions
  • Friction and damping
  • Wheel-ground interaction
  • Sensor behavior (LiDAR, camera, IMU)
  • Actuator dynamics

In short:

URDF defines the robot. Gazebo defines how it behaves.

Extending URDF for Gazebo

Gazebo uses special <gazebo> tags to augment URDF with physics properties.

Collision and Friction

<gazebo reference="wheel_link">
  <mu1>1.2</mu1>
  <mu2>1.2</mu2>
  <kp>100000</kp>
  <kd>1.0</kd>
</gazebo>

This controls:

  • Wheel traction
  • Slip behavior
  • Contact stiffness

Correct friction values are critical for:

  • Wheel odometry accuracy
  • Controller stability
  • Realistic braking behavior

Mass Distribution & Stability

Incorrect inertia values are one of the most common causes of unstable simulations.

Symptoms include:

  • Robot shaking
  • Wheels bouncing
  • Exploding models

Best practices:

  • Compute inertia from CAD tools when possible
  • Use realistic mass values
  • Keep center of mass low for mobile robots

Sensors in Gazebo

Gazebo allows sensors to be attached directly to URDF links.

Example: IMU

<gazebo reference="imu_link">
  <sensor type="imu" name="imu_sensor">
    <update_rate>200</update_rate>
    <imu>
      <noise>
        <type>gaussian</type>
        <stddev>0.001</stddev>
      </noise>
    </imu>
  </sensor>
</gazebo>

This enables:

  • IMU data for state estimation
  • Sensor fusion testing
  • SLAM validation

Similar plugins exist for:

  • LiDAR
  • Cameras
  • GPS
  • Depth sensors

URDF, Gazebo, and Control Systems

URDF also defines transmissions, which connect joints to controllers.

<transmission name="wheel_trans">
  <type>transmission_interface/SimpleTransmission</type>
  <joint name="wheel_joint">
    <hardwareInterface>hardware_interface/VelocityJointInterface</hardwareInterface>
  </joint>
</transmission>

This enables:

  • PID velocity control
  • Differential drive controllers
  • Ackermann steering models
  • Model Predictive Control (MPC) testing

Typical Real-World Workflow

  1. Design robot structure in URDF/Xacro
  2. Attach collision & inertial models
  3. Add Gazebo physics parameters
  4. Integrate sensors
  5. Connect controllers
  6. Validate in Gazebo
  7. Deploy same URDF to real hardware

A well-built URDF often transitions to hardware with minimal changes.

Common Pitfalls

  • Overly complex collision meshes
  • Incorrect inertia tensors
  • Unrealistic friction values
  • Misaligned coordinate frames
  • Treating simulation as “just visuals”

Simulation is only useful if it is physically honest.

Why This Matters

Gazebo + URDF enable:

  • Faster iteration
  • Safer testing
  • Reproducible experiments
  • Hardware-agnostic development

For mobile robots, ADAS systems, and autonomous vehicles, simulation is not optional—it is foundational.

  1. Introduction & Robot Description (URDF)

Learn the basics of robot modeling and setting up your initial file structure.

  1. Building the Links & Joints (Xacro)

Converting raw XML into clean, modular Xacro code.

  1. Gazebo Integration & Physics

Bringing the model into the simulator and adding collision/inertial properties.

  1. Differential Drive & Movement

Using the Gazebo ROS plugin to make the robot move via /cmd_vel.

  1. Adding Sensors (LiDAR)

Simulating a laser scanner to perceive the environment.

  1. Adding Sensors (Cameras)

Integrating visual data and camera plugins into the simulation.

<
Previous Post
“Go Here, Then Here, Then Here”: Automating Robot Patrols with follow_waypoints
📄
CV
View Resume
>
Next Post
🚗 Building an Ackermann-Drive Autonomous Vehicle Simulator in ROS & Gazebo