Skip to content

Software Installation#

This guide provides the complete installation workflow for AutoSDV. Follow these steps sequentially to set up a fully functional autonomous vehicle software stack.

System Requirements#

Choose one of the following platforms:

  1. NVIDIA Jetson AGX Orin 64GB (Recommended for vehicle deployment)
  2. Ubuntu 22.04 PC with NVIDIA GPU (For development and testing)
  3. Docker Environment (For simulation and development)

Storage Requirements#

  • Minimum 256GB SSD (NVMe recommended for Jetson)
  • At least 100GB free space for software installation

Installation Methods#

Choose the installation method that best fits your needs:

Method Best For Difficulty Customization
Automatic Setup Most users, production deployment Easy Limited
Manual Setup Advanced users, custom configurations Advanced Full
Docker Setup Development, testing, simulation Easy Limited

Installation Workflow#

Step 1: Prepare Operating System#

For NVIDIA Jetson AGX Orin#

  1. Download and install NVIDIA SDK Manager
  2. Flash the Jetson with the following configuration:
    • JetPack SDK version: 6.0 (exactly - not 6.1 or 6.2)
    • Install all CUDA and TensorRT packages
    • Flash to external NVMe SSD (not internal eMMC)

For Ubuntu 22.04 PC#

  1. Install Ubuntu 22.04 LTS
  2. Install NVIDIA drivers (version 550 or higher): 
    sudo apt update
sudo apt install nvidia-driver-550
  3. Install CUDA 12.3 using the deb (network) installer
  4. Install TensorRT 8.6 GA

For Docker#

See Docker Setup for containerized installation (skips Steps 1-3).

Step 2: Install ZED SDK (if using ZED camera)#

ZED SDK must be installed manually before proceeding.

The ZED SDK and ZED Link drivers are required if you're using ZED cameras. See ZED SDK Installation Guide for detailed instructions.

Note: This is a manual installation step. The automated setup script does NOT install ZED SDK.

Step 3: Install AutoSDV Software#

Choose your installation method:

The automated setup script installs all dependencies and configures your system.

  1. Clone the repository: 

    cd ~
git clone https://github.com/NEWSLabNTU/AutoSDV.git
cd AutoSDV

  2. Run the automated setup: 

    ./setup.sh

This script automatically installs: - ROS 2 Humble - Autoware 2025.02 - Blickfeld Scanner Library (for Cube1 LiDAR, with license acceptance) - Velodyne drivers (via rosdep) - NMEA/serial drivers (via rosdep) - All other ROS dependencies

The script will prompt you for: - Autoware Debian packages: Install pre-built packages or build from source - CycloneDDS kernel buffers: Configure system-wide network buffers (recommended) - Blickfeld license: Accept license terms for Cube1 LiDAR support

Note: ZED SDK is NOT installed by this script and must be installed manually beforehand.

  1. Install and configure direnv:

    AutoSDV uses direnv for automatic environment activation:

    # Install direnv
sudo apt install direnv

# Add to your shell (bash)
echo 'eval "$(direnv hook bash)"' >> ~/.bashrc
source ~/.bashrc

# Allow .envrc in AutoSDV directory
cd ~/AutoSDV
direnv allow

    After this, the ROS 2 and Autoware environments will automatically activate when you enter the AutoSDV directory.

Manual Environment Setup (Advanced)#

Manually install ROS 2, Autoware, and dependencies for full control:

See Manual Environment Setup for detailed instructions

Docker Setup (Alternative)#

Run AutoSDV in a containerized environment:

See Docker Setup for detailed instructions

Step 4: Build and Verify#

After installing AutoSDV software, build the project:

cd ~/AutoSDV
make prepare  # Install ROS dependencies
make build    # Build the project

Verify the installation by launching the system:

source install/setup.bash
ros2 launch autosdv_launch autosdv.launch.yaml

If successful, you should see the system starting without critical errors.

Troubleshooting#

Build Errors#

If you encounter build errors:

# Clean and rebuild
./clean_build.sh clean
make build

Missing Dependencies#

# Update rosdep database
rosdep update

# Install missing dependencies
cd ~/AutoSDV
rosdep install --from-paths src --ignore-src -r -y

CUDA/TensorRT Issues#

Verify CUDA installation: 

nvcc --version  # Should show 12.3 or compatible
nvidia-smi      # Should show driver 550+

ROS 2 Environment Issues#

Always source ROS 2 before building: 

source /opt/ros/humble/setup.bash

Next Steps#

After successful installation:

Getting Help#

If you encounter issues not covered here: - Check the AutoSDV GitHub Issues - Review the troubleshooting sections in each installation method page