How to Install Isaac ROS 5.0 on Ubuntu 24.04 with Docker

Isaac ROS 5.0 arrived on September 21, 2026 with ROS 2 Lyrical Luth as its supported ROS distribution and a new Ubuntu 24.04 package ecosystem. If you have an NVIDIA GPU workstation, the cleanest way to start is Docker: NVIDIA lists Docker as the recommended isolation mode because Isaac ROS dependencies stay inside the container instead of modifying the host system. This tutorial walks through the supported x86_64 setup and gets the Isaac ROS command-line environment ready for your first workspace.

Check your hardware before installing Isaac ROS 5.0

Isaac ROS 5.0 does not support every Ubuntu and NVIDIA configuration. NVIDIA's tested x86_64 combination is Ubuntu 24.04 with an NVIDIA GPU based on the Ampere architecture or newer, at least 8 GB of RAM, CUDA 13.2 or newer, NVIDIA driver 595 or newer, and at least 32 GB of available disk space. NVIDIA also supports Jetson Thor, Jetson Orin with JetPack 7.2, and DGX Spark, but those require different setup paths from the x86 workstation procedure used here.

Check the operating system first:

lsb_release -a

You should see Ubuntu 24.04. Then check the NVIDIA driver and GPU:

nvidia-smi

The output should show an NVIDIA GPU and a driver version of at least 595. If the command is missing or the GPU is not detected, stop here and fix the host NVIDIA driver before continuing. Isaac ROS uses NVIDIA acceleration inside its development environment, so a working host GPU stack is a prerequisite rather than an optional optimization.

Use Docker instead of installing Isaac ROS directly on Ubuntu

Isaac ROS 5.0 offers Docker, a virtual environment, and bare-metal installation modes. NVIDIA recommends Docker for most users because Debian packages and Python packages are installed inside the container, leaving the host environment largely untouched. That matters for robotics development because packages such as CUDA, OpenCV, TensorRT, and ROS components can have version requirements that conflict with software you already use.

The release documentation says Isaac ROS 5.0 has been tested with CUDA 13.2 on x86 systems, OpenCV 4.6.0, and TensorRT 10.16.*. Installing different versions directly onto the host can therefore create dependency conflicts. Docker gives you a cleaner boundary while still allowing the container to access the NVIDIA GPU.

Create an Isaac ROS workspace

Isaac ROS uses a workspace directory for source packages, configuration, and development files. Create the workspace and its source directory before initializing the development environment.

mkdir -p ~/workspaces/isaac_ros-dev/src

echo 'export ISAAC_ROS_WS="${ISAAC_ROS_WS:-${HOME}/workspaces/isaac_ros-dev/}"' >> ~/.bashrc

source ~/.bashrc

The ISAAC_ROS_WS environment variable now points to the workspace. Keeping this path in your shell configuration means later Isaac ROS commands can find the workspace without requiring you to type its full location each time.

Add the Isaac ROS 5.0 package repository

For an x86_64 Ubuntu 24.04 machine, NVIDIA provides a dedicated Isaac ROS repository for release 5.0. Pinning the repository to release-5.0 is useful when following a tutorial written specifically for this version because it avoids silently moving to a later minor release.

k="/usr/share/keyrings/nvidia-isaac-ros.gpg"

curl -fsSL [https://isaac.download.nvidia.com/isaac-ros/repos.key](https://isaac.download.nvidia.com/isaac-ros/repos.key) | 
sudo gpg --dearmor | sudo tee -a "$k" > /dev/null

f="/etc/apt/sources.list.d/nvidia-isaac-ros.list"

sudo touch "$f"

s="deb [signed-by=$k] [https://isaac.download.nvidia.com/isaac-ros/release-5.0](https://isaac.download.nvidia.com/isaac-ros/release-5.0) noble main"

grep -qxF "$s" "$f" || echo "$s" | sudo tee -a "$f"

sudo apt-get update

The repository now gives Ubuntu's package manager access to the Isaac ROS 5.0 packages, including the Isaac ROS command-line interface. Using the version-specific repository also makes the setup reproducible: another machine following the same instructions starts from the same Isaac ROS release line.

Install the Isaac ROS command-line tool

Install the Isaac ROS command-line interface from the repository you just configured.

sudo apt-get install isaac-ros-cli

The command-line tool is the part that prepares and manages the Isaac ROS development environment. You do not need to install the complete Isaac ROS software stack directly onto Ubuntu when using Docker mode.

Install Docker and NVIDIA Container Toolkit

The Docker workflow requires a working Docker Engine and the NVIDIA Container Toolkit. The toolkit allows containers to access the NVIDIA GPU instead of running as ordinary CPU-only containers. NVIDIA's Isaac ROS instructions also recommend adding your user to the Docker group so that you can use Docker without prefixing every command with sudo.

After installing Docker Engine and the NVIDIA Container Toolkit using their respective Ubuntu installation procedures, add your account to the Docker group:

sudo usermod -aG docker $USER
newgrp docker

Restart Docker before running the Isaac ROS checks:

sudo systemctl daemon-reload
sudo systemctl restart docker

At this point the host has Docker, the NVIDIA container runtime, and the Isaac ROS package repository. The next checks are deliberately simple because it is easier to fix a runtime problem before Isaac ROS is involved than after the development container has been created.

Verify Docker can access the NVIDIA GPU

First confirm that Docker sees its configured runtimes:

docker info | grep -E "Runtimes|Default Runtime"

Then make sure an ordinary container can start:

docker run --rm hello-world

Finally, test GPU access directly:

docker run --rm --gpus all ubuntu:24.04 bash -lc 'echo "NVIDIA runtime OK"'

The final command should print NVIDIA runtime OK. If it fails with an NVIDIA runtime or GPU-related error, do not continue with Isaac ROS yet. The problem is in the Docker or NVIDIA Container Toolkit configuration, and Isaac ROS will not be able to repair that missing host capability for you.

Initialize Isaac ROS in Docker mode

Once Docker and GPU access pass their checks, initialize the Isaac ROS development environment:

sudo isaac-ros init docker

This selects Docker as the Isaac ROS isolation mode and prepares the environment for subsequent development commands. NVIDIA describes Docker as the highest-isolation option in its Isaac ROS 5.0 setup documentation, while virtual-environment mode provides less isolation and bare-metal mode installs dependencies directly into the host system.

If you need extra dependencies for a particular camera, sensor, or robotics package, do not immediately install those packages into Ubuntu. Isaac ROS provides a Docker image configuration mechanism for extending the development environment while retaining the container boundary.

Open the Isaac ROS development environment

With the workspace and Docker environment initialized, you can move on to an Isaac ROS package tutorial. The workspace created earlier gives the development environment a predictable location for source packages, while Docker supplies the tested dependency environment.

For example, verify that the workspace variable is available:

echo "$ISAAC_ROS_WS"

You should get a path ending in workspaces/isaac_ros-dev/. The src directory inside it is where Isaac ROS source repositories can be placed when a particular package tutorial requires building from source.

Understand what changed in Isaac ROS 5.0 before adding packages

There is a significant version boundary here. Isaac ROS 5.0 moved to ROS 2 Lyrical Luth, the ROS 2 release introduced in May 2026 and supported as a Long Term Support release through May 2031. That means instructions written for older Isaac ROS releases should not automatically be copied into a 5.0 environment, particularly when they install ROS packages or interact directly with the former NITROS interfaces.

Isaac ROS 5.0 also rebuilt NITROS around ROS 2 Lyrical's rosidl::Buffer and removed several older NITROS packages. NVIDIA says code that directly calls the removed NITROS APIs or types requires source-level migration. If an older robotics project fails after moving to 5.0, checking whether it depends on those removed interfaces should be one of the first troubleshooting steps.

Know when this setup is not the right one

This tutorial targets the officially supported x86_64 Ubuntu configuration. It is not the procedure for a Jetson board or DGX Spark system. NVIDIA lists Jetson Thor and Jetson Orin with JetPack 7.2 separately, and the Jetson workflow has platform-specific Docker and hardware setup requirements.

The same caution applies to unsupported host combinations. A newer or older Ubuntu release, an older NVIDIA architecture, an older driver, or a different CUDA stack may still work in some circumstances, but NVIDIA does not list those combinations as the tested and officially supported Isaac ROS 5.0 configuration. For a development machine where reproducibility matters, staying inside the documented matrix is the simpler starting point.

Once the GPU runtime passes its checks and isaac-ros init docker completes, the difficult part of the installation is done. From there, the useful next step is to choose one Isaac ROS package—such as perception, visual simultaneous localization and mapping, navigation, or manipulation—and follow its 5.0 tutorial inside the prepared workspace rather than mixing commands from older releases.