Gazebo to Isaac Sim The Complete ROS 2 Migration

Migrate a ROS 2 robot from Gazebo to Isaac Sim: URDF import, A/B measurement, PhysX drives, sensors, Nav2 proof
What you'll learn

---

Import a working Gazebo URDF robot into Isaac Sim with a scripted, reproducible importer run, and read back every link and joint to verify what transferred
Build an A/B rig that drives the same robot in Gazebo and Isaac Sim from one command stream, and prove equivalence with numeric diffs against stated tolerance
Diagnose a divergence from the shape of its plot: flat means frames, linear means a rate, steep-then-flat means a limit, oscillating geometry, a step an event
Trace a real migration failure to its mechanism, such as a URDF damping value that becomes PhysX's velocity tracking gain, and fix it with measured confirmation
Requirements

---

A working ROS 2 robot in Gazebo, or willingness to build the course's one from scratch in Section 1. You should be comfortable with a Linux terminal, Python and URDF/xacro.
Ubuntu 24.04 with ROS 2 Jazzy and Gazebo Harmonic. Isaac Sim 6.0.1 needs an NVIDIA RTX GPU; 6 GB VRAM is enough to run both simulators side by side.
No Isaac Sim or Omniverse experience required
Description

---

This course contains the use of artificial intelligence.
Your robot already works in Gazebo. You have an SDF or Xacro, tuned controllers, and a Nav2 configuration that took weeks to get right. Now you need to move it to NVIDIA Isaac Sim for photoreal sensors, GPU physics, synthetic data, or Omniverse workflows.
The problem is that almost nobody teaches the migration itself.
What this course teaches
You will migrate a working Gazebo robot to Isaac Sim andprove with measurements that it behaves the same, or identify exactly why it does not.
The core of the course is anA/B simulation rig
Same robot
Same ROS 2 nodes
Same commands
Gazebo and Isaac Sim running side by side
Behaviour compared using real numerical data
We begin with a migration that fails. The robot drives almost correctly but rotates only half as far as expected. We test friction, caster behaviour, joints, and PhysX settings until the real cause is isolated.
What you will investigate
Geometry, mass, and inertia
Joints and drive configuration
Contact and friction
Cameras and sensors
ROS 2 interfaces
Controllers and Nav2
Full Gazebo to Isaac Sim behaviour
You will also learn to read divergence plots and recognize patterns such as rate errors, limits, frame offsets, geometry problems, and discrete events.
The course uses
ROS 2 Jazzy
Gazebo Harmonic
NVIDIA Isaac Sim 6.0.1
Every important result comes from reproducible scripts included in the repository. Failures are deliberately created, measured, debugged, and fixed on camera, including cases where the first hypothesis turns out to be wrong.
The goal is not just to import your robot into Isaac Sim. It is to make surethe migrated robot actually behaves like the original.
Who this course is for

---

ROS 2 roboticists with a robot already working in Gazebo who need it in Isaac Sim, and need proof it still behaves the same rather than a hope that it does.
Simulation and robotics engineers told to move to Omniverse for photoreal sensors, GPU-scale physics or synthetic data, who own the risk if a tuned stack breaks.
Anyone who has tried the URDF importer, got a robot that looked right and behaved differently, and could not find where the difference came from.
Homepage

---