Installation Guide¶

Quick Install¶

The fastest way to install all required components is via the provided scripts:

# Install everything (Python package + NS-3 + Mininet + Gazebo/ROS 2)
make install

# Or install only what you need:
make install-python   # Python package only
make install-ns3      # NS-3 3.38 + NR v2.4 + CORNET patches (stable default)
make install-ns3-v47  # NS-3 3.47 + NR v4.2 (explicit stepping stone)
make install-ns3-v51  # NS-3 3.48 + NR v5.1 (explicit; pins in scripts/patches/ns3/v5.1-ns3.48/README.md)
make install-mininet  # Mininet-WiFi + Docker
make install-gazebo   # ROS 2 Humble + Gazebo Classic 11 + TurtleBot3
make install-px4      # PX4 Classic iris SITL (aerial pack; not part of `make install`)

# Verify all components are functional:
make verify

All scripts are in scripts/install/ and are idempotent — safe to re-run.

Docker alternative: If you prefer a containerised setup, Docker images are available under scripts/docker/. Run docker compose up from that directory to start a pre-configured CORNET environment without manual dependency installation.

Python Package¶

pip install cornet-framework
# or from source
git clone https://github.com/rbccps-iisc/CORNET
cd CORNET
pip install -e .[dev]

Requires Python 3.10+.

System Prerequisites¶

CORNET plugins require system-level simulators. Install only what you need for your chosen backend.

NS-3 with 5G NR (for `network.plugin: ns3`)¶

Version note: CORNET's default install targets NS-3 3.38 + NR v2.4. Two explicit lanes sit beside it: NS-3 3.47 + NR v4.2 (make install-ns3-v47) and NS-3 3.48 + NR v5.1 (make install-ns3-v51). Neither replaces the default. See scripts/patches/ns3/CAPABILITY_MATRIX.yaml for what each lane has validated.

On Ubuntu 22.04 the v5.1 lane builds with g++ 11 (C++20), Python 3.10, and Ninja 1.10. NS-3 3.48 requires CMake 3.25 or newer. Ubuntu 22.04's apt package is 3.22, which stops ./ns3 configure. This machine has g++ 11.4.0, python3 3.10.12, and ninja 1.10.1; its system CMake is 3.22.1, so the v5.1 install needs a newer CMake on PATH. Without sudo, install one for this user and leave ~/.local/bin ahead of /usr/bin:

python3 -m pip install --user 'cmake>=3.25'
export PATH="$HOME/.local/bin:$PATH"

make install-ns3-v51 prefers $HOME/.local/bin/cmake when that file exists. NS-3 3.48 is pinned to d2add90b452d600cfb4859baed8e9ea633519447 and 5G-LENA v5.1 to cedceadda17392c90587fb9400eb9b1f8c236713.

# Stable default: NS-3 3.38 + NR v2.4 into ~/ns-3-dev
make install-ns3

# Stepping stone: NS-3 3.47 + NR v4.2 into ~/ns-3-dev-v47
make install-ns3-v47

# Latest explicit lane: NS-3 3.48 + NR v5.1 into ~/ns-3-dev-v51
# Needs CMake >= 3.25 on PATH (see above). Does not replace the default.
make install-ns3-v51

The installer clones the pinned tag, checks the commit, configures, builds, then applies the lane's profiles/full.txt. A manual checkout should use those same pins from scripts/patches/ns3/v5.1-ns3.48/README.md (or the v2.4 / v4.2 READMEs) rather than a floating branch.

Run the compat check script before applying patches to a pre-existing installation:

# Default check (v4.2-ns3.47)
python3 scripts/check_ns3_compat.py --ns3-dir /path/to/ns-3-dev

# Legacy check (v2.4-ns3.38)
python3 scripts/check_ns3_compat.py --ns3-dir /path/to/ns-3-dev --patch-set v2.4-ns3.38

The ns3 plugin searches for the NS-3 build at NS3_DIR environment variable, then ~/ns-3-dev/.

Mininet-WiFi + Docker (for `network.plugin: mininet`)¶

# Install Mininet-WiFi
git clone https://github.com/intrig-unicamp/mininet-wifi
cd mininet-wifi
sudo util/install.sh -Wlnfv

# Install Docker (for container nodes)
curl -fsSL https://get.docker.com | sh
sudo usermod -aG docker $USER

# Mininet requires root — always run with sudo:
sudo python -m cornet tasks/<name>

Gazebo Classic 11 + ROS 2 Humble (for `robot.plugin: gazebo`)¶

# ROS 2 Humble (Ubuntu 22.04)
sudo apt-get install -y ros-humble-desktop ros-humble-gazebo-ros-pkgs

# Source ROS 2 in every terminal (or add to ~/.bashrc)
source /opt/ros/humble/setup.bash

The gazebo plugin calls ros2 launch internally; ROS 2 must be sourced before running.

make install-gazebo also installs the TurtleBot3 burger model that catalogue compose uses:

/opt/ros/humble/share/turtlebot3_gazebo/models/turtlebot3_burger/model.sdf

Gazebo load benchmarks (python -m cornet bench gazebo) spawn that model. Source ROS 2 before those runs.

PX4 Classic iris (for the `px4_x500` catalogue pack)¶

make install does not build PX4. After Gazebo is installed:

make install-px4

The script clones ~/simulation/PX4-Autopilot at 36006b6d703a421175587d386a535bbdf8eb0a9c with the Gazebo Classic submodule 5b6966ed572a02e8273f446acb504a45a841ca53, then runs DONT_RUN=1 make px4_sitl gazebo-classic. The airframe is iris. It also installs libgz-transport13, libgz-msgs10, and libgz-utils2 from the OSRF Gazebo apt repository, because that binary links them. An existing checkout at a different commit is left in place.

Warehouse worlds are optional. Set CORNET_SMALL_WAREHOUSE and CORNET_LARGE_WAREHOUSE to a world file or package directory. When unset, the suite looks at the known local AWS and ARTPARK checkouts, then at vendored cornet/catalog/worlds/ paths. Missing worlds are skipped.

Git LFS¶

Catalogue meshes and textures (cornet/catalog/**/*.dae, *.stl, *.png, *.jpg) are stored with Git LFS. After cloning, install the client once and pull the objects:

git lfs install
git lfs pull

The scenario compiler refuses to run when a vendored asset is still a Git LFS pointer. On this machine the client is ~/.local/bin/git-lfs 3.8.0, because installing the apt package needs sudo.

Verify Installation¶

python -m cornet --help
python -m pytest tests/ -v