Gazebo

Gazebo is an open source 3D robotics simulator that models robots, sensors, and environments with physics, so developers can test software and generate sensor data before running it on real hardware. It is stewarded by Open Robotics alongside ROS, governed by the Open Source Robotics Alliance, and connects to ROS 2 through the ros_gz bridge.

What is Gazebo?

Gazebo is the simulator most closely tied to ROS. It loads a robot's model, often from a URDF or SDF file, places it in a 3D world, and simulates physics, contact, and sensors such as cameras, depth cameras, LiDAR, and IMUs. Robot software can then run against the simulated robot the same way it would against a real one.
Gazebo began in 2002 at the University of Southern California and is now stewarded by Open Robotics. Since 2024, its development has been governed by the Open Source Robotics Alliance, which also oversees ROS, Open-RMF, and other projects and is funded by paid memberships. The original simulator, now called Gazebo Classic, reached end of life in January 2025. Its successor, first released as Ignition and renamed Gazebo in 2022, ships long-term support releases such as Fortress, Harmonic, and Jetty. You will also see it called Gazebo Sim, gz sim after its command, or Ignition Gazebo in older docs.

Key takeaways

  • Gazebo simulates robots, sensors, and physics so software can be tested before it runs on hardware.
  • It connects to ROS 2 through the ros_gz bridge and is governed, with ROS, by the Open Source Robotics Alliance.
  • Gazebo Classic reached end of life in January 2025, and current releases carry names such as Harmonic and Jetty.

How it works

A Gazebo world is described in SDF, the Simulation Description Format, which lists models, lights, physics settings, and plugins. A physics engine steps the world forward in time and resolves contacts, and sensor plugins render what each simulated sensor would see. The ros_gz bridge passes messages between Gazebo and ROS 2, so the same ROS nodes that drive a real robot can subscribe to simulated camera images and publish motor commands. RViz is often open at the same time, but it only visualizes. Gazebo is the one doing the simulating.
Gazebo produces the simulated data, and RViz only displays it.

Why it matters

Simulation lets teams test a navigation stack or a grasping pipeline thousands of times without breaking hardware, and it can produce labeled sensor data at low cost, since the simulator knows the ground truth. The catch is the gap between simulated and real sensors: rendered images, perfect depth, and idealized contact rarely match what a real robot records. Comparing simulated and real recordings side by side is how teams find that gap before it shows up as a failed sim-to-real transfer.
Simulation lets teams test robot software before it touches hardware.

Frequently asked questions

What is the difference between Gazebo and RViz?

Gazebo simulates a robot and its world, including physics and sensors. RViz visualizes data that a real or simulated robot publishes. Teams often run both: Gazebo produces the data, and RViz displays it.

What is Gazebo Classic?

Gazebo Classic is the original Gazebo simulator, last released as Gazebo 11. It reached end of life in January 2025, and projects are expected to move to the newer Gazebo releases.

Is Gazebo free?

Yes. Gazebo is open source under the Apache 2.0 license.
black and white photo of Jesse Mostipak
Director of Growth
Jesse Mostipak is the Director of Growth at Voxel51, where the work is helping humans find and trust what the brand knows, and teaching the Google knowledge graph and the LLMs answering on their behalf to do the same. That question, how knowledge gets built inside a system, is one Jesse has been chasing for years. Earlier versions of it ran through a New York City high school science classroom, data science and machine learning, and developer relations at Kaggle, Posit (formerly RStudio), and Baseten. The answer doesn't change much depending on whether the learner is a teenager, a software engineer, or a knowledge graph. Jesse holds a Master's in Education from CUNY Hunter College. LinkedIn
See all articles by Jesse Mostipak
Last updated November 19, 2026

Building visual or physical AI?

Let's talk.