Setup Jetson Xavier NX

Recommended Versions for Jetson Xavier NX (as of 2025)

Best Choice: JetPack 5.1.4

• JetPack: 5.1.4 (Latest stable for Xavier NX)
• L4T: 35.6.0
• CUDA: 11.4
• TensorRT: 8.5.2
• cuDNN: 8.6.0
• Ubuntu: 20.04 LTS

Where to Find Official NVIDIA Resources

1. Main Starting Points

• JetPack Hub: https://developer.nvidia.com/embedded/jetpack
• Jetson Download Center: https://developer.nvidia.com/embedded/downloads
• SDK Manager: https://developer.nvidia.com/sdk-manager

2. Direct Navigation Path

NVIDIA Developer → Embedded Computing → Jetson → JetPack SDK
                                         ↓
                               Select your device (Xavier NX)
                                         ↓
                               Download SD Card Image or SDK Manager

3. Essential Tools

For SD Card Setup (Easiest)

• BalenaEtcher: https://www.balena.io/etcher/
• SD Card Formatter: https://www.sdcard.org/downloads/formatter/

For Advanced Setup

• NVIDIA SDK Manager: For full control over installation
• L4T BSP: For custom configurations

Step-by-Step Guide to Find Resources

Step 1: Check Your Current Version

# Check JetPack version
sudo apt show nvidia-jetpack

# Check L4T version
cat /etc/nv_tegra_release

# Check CUDA version
nvcc --version

Step 2: Find Compatible Software

Official NVIDIA Resources

1. JetPack Archive: https://developer.nvidia.com/embedded/jetpack-archive
2. Jetson Linux Archive: https://developer.nvidia.com/embedded/linux-tegra-archive
3. CUDA Zone: https://developer.nvidia.com/cuda-zone

Documentation

• Xavier NX Getting Started: https://developer.nvidia.com/embedded/learn/get-started-jetson-xavier-nx-devkit
• JetPack Documentation: https://docs.nvidia.com/jetson/

Step 3: Download Checklist

For Fresh Installation

1. SD Card Image (easiest method)
   • Size: ~6-7GB compressed
   • Filename pattern: jetson-xavier-nx-jp514-sd-card-image.zip
2. SDK Manager (for custom installation)
   • Requires Ubuntu 18.04/20.04 host PC
   • Can install specific components

For Development

# After installation, get development tools
sudo apt update
sudo apt install nvidia-jetpack-dev

# For AI/ML development
sudo apt install python3-pip
pip3 install numpy tensorflow torch torchvision

Version Compatibility Matrix

Important Notes

⚠️ Version Warnings

• Never use JetPack 6.x – It’s ONLY for Orin series
• Stick with JetPack 5.1.x for best Xavier NX support
• JetPack 4.6.x – Still supported but older Ubuntu 18.04

💡 Best Practices

1. Always verify image checksums when available
2. Use quality SD cards (SanDisk Extreme, Samsung EVO)
3. Keep backups of working configurations
4. Document your setup process

🔧 Troubleshooting Resources

• NVIDIA Forums: https://forums.developer.nvidia.com/c/agx-autonomous-machines/jetson-embedded-systems/
• JetsonHacks: https://jetsonhacks.com/
• GitHub Issues: Search for specific error messages

Quick Setup Commands After Fresh Install

# Update system
sudo apt update && sudo apt upgrade

# Install JetPack components
sudo apt install nvidia-jetpack

# Check installation
sudo jetson_clocks --show
sudo tegrastats

# Install useful tools
sudo apt install htop jtop nano git build-essential

Complete ORB-SLAM3 Setup Guide for Jetson Xavier NX with RealSense D455

System Requirements

• Hardware: NVIDIA Jetson Xavier NX (8GB RAM, 32GB storage)
• OS: Ubuntu 20.04.6 LTS
• Camera: Intel RealSense D455

Table of Contents

1. System Preparation
2. ROS2 Foxy Installation
3. RealSense SDK Installation
4. ORB-SLAM3 Dependencies
5. Pangolin Installation
6. ORB-SLAM3 Installation
7. Testing and Demo
8. Troubleshooting Guide

1. System Preparation

Update System

sudo apt update && sudo apt upgrade -y
sudo apt install -y curl wget git cmake build-essential pkg-config

Check Available Space

df -h
# Ensure you have at least 5GB free space

Optimize Jetson Performance

# Set maximum performance mode
sudo nvpmodel -m 0
sudo jetson_clocks

# Create optimization script
cat > ~/optimize_jetson.sh << 'EOF'
#!/bin/bash
echo "Optimizing Jetson Xavier NX..."
sudo nvpmodel -m 0
sudo jetson_clocks
echo 3 | sudo tee /proc/sys/vm/drop_caches
echo "Optimization complete!"
EOF

chmod +x ~/optimize_jetson.sh

2. ROS2 Foxy Installation

Add ROS2 Repository

# Install required packages
sudo apt install -y software-properties-common
sudo add-apt-repository universe

# Add ROS2 GPG key
sudo curl -sSL https://raw.githubusercontent.com/ros/rosdistro/master/ros.key -o /usr/share/keyrings/ros-archive-keyring.gpg

# Add repository
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/ros-archive-keyring.gpg] http://packages.ros.org/ros2/ubuntu $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/ros2.list > /dev/null

Install ROS2 Foxy

sudo apt update
sudo apt install -y ros-foxy-desktop python3-argcomplete

# Install additional tools
sudo apt install -y \
  python3-colcon-common-extensions \
  python3-rosdep \
  python3-vcstool

# Initialize rosdep
sudo rosdep init
rosdep update

# Add to bashrc
echo "source /opt/ros/foxy/setup.bash" >> ~/.bashrc
source ~/.bashrc

Install ROS2 Development Tools

sudo apt install -y \
  ros-foxy-cv-bridge \
  ros-foxy-image-transport \
  ros-foxy-diagnostic-updater \
  ros-foxy-diagnostic-msgs \
  ros-foxy-tf2-ros \
  ros-foxy-tf2-tools \
  ros-foxy-tf2-sensor-msgs \
  ros-foxy-xacro \
  ros-foxy-joint-state-publisher \
  ros-foxy-robot-state-publisher

3. RealSense SDK Installation

Build from Source (Required for ARM64)

# Install dependencies
sudo apt-get install -y git libssl-dev libusb-1.0-0-dev pkg-config libgtk-3-dev
sudo apt-get install -y libglfw3-dev libgl1-mesa-dev libglu1-mesa-dev

# Clone librealsense
cd ~
git clone https://github.com/IntelRealSense/librealsense.git
cd librealsense

# Setup udev rules
sudo cp config/99-realsense-libusb.rules /etc/udev/rules.d/
sudo udevadm control --reload-rules && sudo udevadm trigger

# Build with CUDA support
mkdir build && cd build
cmake .. \
    -DBUILD_EXAMPLES=true \
    -DCMAKE_BUILD_TYPE=Release \
    -DFORCE_RSUSB_BACKEND=true \
    -DBUILD_WITH_CUDA=true \
    -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc

make -j2
sudo make install
sudo ldconfig

Install ROS2 RealSense Wrapper

# Create workspace
mkdir -p ~/ros2_ws/src
cd ~/ros2_ws/src

# Clone RealSense ROS2 wrapper
git clone https://github.com/IntelRealSense/realsense-ros.git -b 4.51.1

# Build
cd ~/ros2_ws
source /opt/ros/foxy/setup.bash
colcon build --symlink-install

# Source workspace
source install/setup.bash
echo "source ~/ros2_ws/install/setup.bash" >> ~/.bashrc

4. ORB-SLAM3 Dependencies

Install System Dependencies

sudo apt install -y \
  libgtk2.0-dev \
  libavcodec-dev libavformat-dev libswscale-dev \
  libtbb2 libtbb-dev \
  libjpeg-dev libpng-dev libtiff-dev \
  libglew-dev libboost-all-dev libssl-dev \
  libeigen3-dev \
  libgl1-mesa-dev libegl1-mesa-dev libwayland-dev \
  libopencv-dev \
  python3-dev python3-numpy python3-pip

5. Pangolin Installation

Install Pangolin with C++14 Support

cd ~
git clone https://github.com/stevenlovegrove/Pangolin.git
cd Pangolin
mkdir build && cd build

# Configure with warnings disabled
cmake .. \
  -DCMAKE_BUILD_TYPE=Release \
  -DBUILD_TOOLS=OFF \
  -DBUILD_EXAMPLES=OFF \
  -DBUILD_TESTS=OFF \
  -DBUILD_PANGOLIN_PYTHON=OFF \
  -DCMAKE_CXX_FLAGS="-w -Wno-error" \
  -DCMAKE_C_FLAGS="-w -Wno-error"

make -j2
sudo make install
sudo ldconfig

Note: If you encounter OpenEXR errors, the -w flag disables all warnings and allows compilation to proceed.

6. ORB-SLAM3 Installation

Clone and Prepare ORB-SLAM3

cd ~
git clone https://github.com/UZ-SLAMLab/ORB_SLAM3.git
cd ORB_SLAM3

# Fix Eigen paths for ARM architecture
chmod +x build.sh
sed -i 's/Eigen\//eigen3\/Eigen\//g' $(find . -type f -name "*.h" -o -name "*.cc" -o -name "*.cpp")

# Fix incorrect unsupported Eigen path
find . -type f \( -name "*.cc" -o -name "*.cpp" -o -name "*.h" \) -exec sed -i 's|unsupported/eigen3/Eigen|unsupported/Eigen|g' {} \;

Build Third-party Libraries

# Build DBoW2
cd Thirdparty/DBoW2
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j2
cd ../../..

# Build g2o
cd Thirdparty/g2o
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j2
cd ../../..

# Build Sophus (optional, skip if errors)
if [ -d "Thirdparty/Sophus" ]; then
    cd Thirdparty/Sophus
    mkdir build && cd build
    cmake .. -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTS=OFF
    make -j2 || echo "Sophus build failed, continuing without it"
    cd ../../..
fi

Extract Vocabulary and Build ORB-SLAM3

# Extract vocabulary
cd Vocabulary
tar -xf ORBvoc.txt.tar.gz
cd ..

# Build ORB_SLAM3 with C++14 support
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_STANDARD=14
make -j2

Create D455 Configuration File

mkdir -p ~/ORB_SLAM3/Examples/RGB-D
cat > ~/ORB_SLAM3/Examples/RGB-D/D455.yaml << 'EOF'
%YAML:1.0

# Camera Parameters
Camera.type: "PinHole"

# Camera calibration parameters
Camera.fx: 382.613
Camera.fy: 382.613
Camera.cx: 320.183
Camera.cy: 236.455

Camera.k1: 0.0
Camera.k2: 0.0
Camera.p1: 0.0
Camera.p2: 0.0

Camera.width: 640
Camera.height: 480
Camera.fps: 30.0

# IR projector baseline times fx
Camera.bf: 40.0

# Color order (0: BGR, 1: RGB)
Camera.RGB: 1

# Depth threshold
ThDepth: 40.0
DepthMapFactor: 1000.0

# ORB Parameters
ORBextractor.nFeatures: 1000
ORBextractor.scaleFactor: 1.2
ORBextractor.nLevels: 8
ORBextractor.iniThFAST: 20
ORBextractor.minThFAST: 7

# Viewer Parameters
Viewer.KeyFrameSize: 0.05
Viewer.KeyFrameLineWidth: 1
Viewer.GraphLineWidth: 0.9
Viewer.PointSize: 2
Viewer.CameraSize: 0.08
Viewer.CameraLineWidth: 3
Viewer.ViewpointX: 0
Viewer.ViewpointY: -0.7
Viewer.ViewpointZ: -1.8
Viewer.ViewpointF: 500
EOF

7. Testing and Demo

Test with Sample Dataset

cd ~/ORB_SLAM3

# Download TUM dataset
wget https://vision.in.tum.de/rgbd/dataset/freiburg1/rgbd_dataset_freiburg1_desk.tgz
tar -xzf rgbd_dataset_freiburg1_desk.tgz

# Run RGB-D example
./Examples/RGB-D/rgbd_tum \
  Vocabulary/ORBvoc.txt \
  Examples/RGB-D/TUM1.yaml \
  rgbd_dataset_freiburg1_desk \
  Examples/RGB-D/associations/fr1_desk.txt

Test with RealSense D455

# Terminal 1: Launch RealSense
source ~/ros2_ws/install/setup.bash
ros2 launch realsense2_camera rs_launch.py \
  depth_module.profile:=640x480x30 \
  rgb_camera.profile:=640x480x30 \
  align_depth.enable:=true

# Terminal 2: Verify camera is working
ros2 topic list
ros2 topic echo /camera/color/image_raw --no-arr

8. Troubleshooting Guide

Error 1: No Space on Device

Symptom: fallocate: fallocate failed: No space left on device

Solution: Clean up disk space

sudo apt-get clean
sudo apt-get autoremove -y
sudo journalctl --vacuum-time=2d
df -h  # Check available space

Error 2: RealSense Package Not Found

Symptom: E: Unable to locate package librealsense2-dkms

Solution: The -dkms package is not available for ARM64. Build from source instead (see Section 3).

Error 3: OpenEXR Compilation Errors

Symptom: Multiple warnings about deprecated functions in Pangolin build

Solution: Disable warnings in CMake

cmake .. -DCMAKE_CXX_FLAGS="-w -Wno-error" -DCMAKE_C_FLAGS="-w -Wno-error"

Error 4: Eigen Path Errors

Symptom: fatal error: unsupported/eigen3/Eigen/MatrixFunctions: No such file or directory

Solution: Fix Eigen paths

find . -type f \( -name "*.cc" -o -name "*.cpp" -o -name "*.h" \) -exec sed -i 's|unsupported/eigen3/Eigen|unsupported/Eigen|g' {} \;

Error 5: C++14 Features Required

Symptom: Errors about std::enable_if_t and other C++14 features

Solution: Build with C++14 support

cmake .. -DCMAKE_CXX_STANDARD=14

Error 6: Memory Issues During Compilation

Symptom: Compilation freezes or crashes

Solution: Use fewer parallel jobs

make -j1  # or make -j2 maximum

Final Verification

After successful installation, you should be able to:

1. Run realsense-viewer and see your D455 camera
2. Run the TUM dataset example and see the 3D reconstruction
3. Launch ROS2 nodes without errors

Important Notes:

• Always run ~/optimize_jetson.sh before intensive SLAM operations
• Move the camera slowly for best tracking results
• Ensure good lighting and textured environments for robust tracking

This guide has been tested on Jetson Xavier NX with Ubuntu 20.04.6 LTS and successfully builds ORB-SLAM3 with RealSense D455 support.