rbonghi / isaac_ros_base

🐋🐋 Multi architecture Isaac ROS for NVIDIA Jetson and x86 - CI based

Isaac ROS Base - Multi architecture & CI based

Multi architecture cross compilable Isaac ROS for x86 and NVIDIA Jetson with Jetpack 5+

Requirements

To build these NVIDIA Docker you can choose one of these options:

1. Desktop
   • NVIDIA Graphic card
2. NVIDIA Jetson
   • Xavier or Orin series
   • NVIDIA Jetpack 5+

Images available

All images are based with:

Software	Version JP=5.0.2	Version JP=5.1
Ubuntu	20.04	20.04
L4T	35.1	35.2
CUDA	11.4	11.4
OpenCV	4.5.0 with CUDA	4.5.0 with CUDA
TensorRT	8.4	8.4
Triton	2.24.0	2.30.0

Images:

Name	AMD64	ARM64	Note	Build Time
rbonghi/isaac_ros_base:humble-core-devel			Minimal packages ros-core and all main package for vision compiled with CUDA	16000s (L4T 35.1) 15000s (L4T 35.2)
rbonghi/isaac_ros_base:humble-base-devel				16000s (L4T 35.1) 16000s (L4T 35.2)
rbonghi/isaac_ros_base:gems-devel			Isaac ROS argus camera is available only on ARM64 platform
rbonghi/isaac_ros_base:isaac-runtime	Soon	Soon

Work with Isaac ROS base

Example to build Isaac ROS packages multi-platform

cd example
docker build -t isaac-ros-base/packages:latest -f Dockerfile.isaac .

Build

There are two ways to use this repository, build locally the isaac_ros_base images or use in CI, but you need to install a local runner on your desktop with NVIDIA graphic card.

Setup Docker runtime on Jetson

To enable access to the CUDA compiler (nvcc) during docker build operations, add "default-runtime": "nvidia" to your /etc/docker/daemon.json configuration file before attempting to build the containers:

{
    "runtimes": {
        "nvidia": {
            "path": "nvidia-container-runtime",
            "runtimeArgs": []
        }
    },
    "default-runtime": "nvidia"
}

You will then want to restart the Docker service or reboot your system before proceeding.

sudo systemctl restart docker.service

Add your user on your docker group

sudo usermod -aG docker $USER

Logout and login on your session.

Build locally

If you want to run locally use and follow the help:

Warning: You can use this script only on:

• x86 machines with NVIDIA graphic card
• NVIDIA Jetson Orin or Xavier series

./docker_build_ros.sh

isaac_ros_runner

Follow README in isaac_ros_runner folder

Multistage images

Name	AMD64	ARM64	Build Time
rbonghi/isaac_ros_base:opencv-4.5.0
rbonghi/isaac_ros_base:devel			2000s (L4T 35.1) 2500s (L4T 35.2)
rbonghi/isaac_ros_base:runtime	Soon	Soon

isaac_ros_base's Issues

Isaac Ros BAse on Jetson Nano

Hi Can i build the ISaac Ros Base on Jetson Nano 4GB

Question about docker build environment: CMake version for docker build seems to be different than my installed version. Environment variable?

Hi, Thankyou for putting together this repository for Docker!
I am attempting to build for Orin Nano Development Kit. I have Jetpack 5.1.1 installed on Orin Nano.
I start a build on Orin Nano using:

cd example
docker build -t isaac-ros-base/packages:latest -f Dockerfile.isaac .\

The build continues, and then fails with:

_Starting >>> isaac_ros_common
Starting >>> isaac_ros_test
--- stderr: isaac_ros_common
CMake Error at CMakeLists.txt:9 (cmake_minimum_required):
CMake 3.23.2 or higher is required. You are running version 3.16.3

Failed <<< isaac_ros_common [0.61s, exited with code 1]
Aborted <<< isaac_ros_test [1.62s]_

I am assuming CMake 3.16.3 is default shipping version with Jetpack 5.1.1?

I will simply update to the latest version of CMake. Is this the correct approach?
Thanks for your help!

[QST] What's the relationship between this project and the official nvidia ngc isaac_ros_deb_base container.

First of all I would say it's a great project to help us create docker images to run issac_ros.

At the same time, there is a official container which is published without dockerfile in ngc catlog, which is https://catalog.ngc.nvidia.com/orgs/nvidia/teams/isaac/containers/ros , and is recommended in https://github.com/NVIDIA-ISAAC-ROS/isaac_ros_common/blob/main/docker/Dockerfile.x86_64.ros2_humble so developers can use it freely(without the dockerfile!).

Does this project have any relationship with the nvidia version of container?

1. Does this project try to reverse-engineering the closed source ngc version?
2. Or it is just a standalone project that will get its own path.

Thanks!