sim4cd - Simulator for a Custom Drone

This package is a simulator tailored for quadcopters with fixed rotors. It offers configurable drone characteristics, allowing users to replicate their own platform. Currently, the simulator integrates seamlessly with PX4 firmware.

sim4cd is a plain Python package. It talks to PX4 over MAVLink (pymavlink) and to its own GUI over ZMQ, so ROS is not required to run it. There is an optional ROS bridge for RViz visualization, described at the end of this file.

The instructions below use ~/sim4cd_ws as the workspace, with PX4 checked out next to sim4cd. Any directory works; only the relative layout matters, and it can be overridden with PX4_DIR.

Docker setup

Clone sim4cd

mkdir -p ~/sim4cd_ws/src
cd ~/sim4cd_ws/src
git clone https://github.com/adrianomcr/sim4cd.git
cd sim4cd

Build the image. It is based on Ubuntu 24.04 and builds PX4 SITL, so the first build takes a while.

docker compose -f docker-compose.yaml build

Give docker access to the X server

xhost +local:docker

Run the Qt GUI

docker compose -f docker-compose.yaml run --rm sim4cd

Or get a shell in the container instead

docker compose -f docker-compose.yaml run --rm sim4cd bash

Native setup

Clone the repo

The Qt GUI lives on the add-python-qt-gui branch.

mkdir -p ~/sim4cd_ws/src
cd ~/sim4cd_ws/src
git clone -b add-python-qt-gui https://github.com/adrianomcr/sim4cd.git

System packages

sudo apt update
sudo apt install -y git build-essential cmake ninja-build \
  python3-venv python3-pip python3-dev \
  libgl1 libglib2.0-0 libxkbcommon-x11-0 \
  libxcb-xinerama0 libxcb-cursor0 libxcb-icccm4 \
  libxcb-image0 libxcb-keysyms1 libxcb-randr0 \
  libxcb-render-util0 libxcb-shape0

On Ubuntu 24.04, install python3.12-venv if python3-venv is not available. You can skip libglib2.0-0 if apt reports it was renamed to libglib2.0-0t64.

Python dependencies (venv)

Ubuntu 23.04 and later block installing pip packages into the system Python (PEP 668). Use a virtualenv:

cd ~/sim4cd_ws/src/sim4cd
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

vtk is pinned to 9.4.1 in requirements.txt (9.6+ rotates the home-tab skybox).

Build PX4 SITL

Clone PX4 next to sim4cd:

cd ~/sim4cd_ws/src
git clone --depth 1 --branch v1.13.3 https://github.com/PX4/PX4-Autopilot.git PX4
git -C PX4 submodule update --depth 1 --init --recursive
cd PX4
make distclean

Install the build-time Python dependencies from PX4's own requirements file. One line in it uses a version specifier that pip 24+ rejects (matplotlib>=3.0.*), so patch it first:

sed -i 's|matplotlib>=3.0\.\*|matplotlib>=3.0|' Tools/setup/requirements.txt

Then, with the sim4cd virtualenv active:

pip install -r Tools/setup/requirements.txt
pip install 'empy==3.3.4'

The virtualenv must stay activated while running make, since the PX4 build shells out to python3 for code generation. Installing into the system Python works too on Ubuntu 20.04/22.04; on 23.04 and later add --user --break-system-packages, which installs into ~/.local without touching any apt-owned file.

The second empy command is required on every distro. v1.13.3 asks for empy>=3.3, which now resolves to 4.x, and 4.x breaks the PX4 uORB code generation with module 'em' has no attribute 'RAW_OPT'.

Build SITL

DONT_RUN=1 make px4_sitl none_iris

On GCC 12 and newer (Ubuntu 22.04 ships GCC 11, 24.04 ships GCC 13), this fails in src/lib/matrix with error: array subscript 1 is above array bounds of 'float [1][1]' [-Werror=array-bounds]. It is a false positive in an unreachable branch of matrix::inv(), fixed in later PX4 releases. Stop v1.13.3 from treating warnings as errors and rebuild:

sed -i 's|^\(\s*\)-Werror$|\1-Wno-error|' cmake/px4_add_common_flags.cmake
DONT_RUN=1 make px4_sitl none_iris

scripts/sim4cd/start_sim.sh finds PX4 by looking for a PX4 or PX4-Autopilot directory next to the repo. If you cloned it somewhere else, point PX4_DIR at it:

export PX4_DIR=/path/to/PX4-Autopilot

Install QGroundControl

QGroundControl is used as the ground control station.

sudo usermod -a -G dialout $USER # Needs to log out and log in
sudo apt-get remove modemmanager -y
sudo apt install gstreamer1.0-plugins-bad gstreamer1.0-libav gstreamer1.0-gl -y
sudo apt install libqt5gui5 -y
sudo apt install libfuse2 -y

Download the AppImage from https://docs.qgroundcontrol.com/master/en/qgc-user-guide/getting_started/download_and_install.html.

cd ~/Downloads
chmod +x ./QGroundControl.AppImage

Qt GUI

The Qt GUI is the main way to load a vehicle config and start the simulator (Start Simulator runs scripts/sim4cd/start_sim.sh).

Run

From scripts/qt_gui, with scripts/ on PYTHONPATH so sim4cd imports resolve:

cd ~/sim4cd_ws/src/sim4cd/scripts/qt_gui
PYTHONPATH=.. python3 qt_sim4cd.py

Alternatively, from the repo root with the venv active:

pip install -e .
sim4cd-gui

Screenshots

Simulate tab: start and stop the simulator and follow the vehicle in the 3D scene

Simulation config: geographic location of the simulation origin and local magnetic field

Simulation config: vehicle dynamics and geometry, with the actuator layout rendered as it is edited

Simulation config: actuator dynamics and curve maps, with the polynomial estimator

Simulation config: full list of simulator parameters

QGroundControl connected to the running simulation, at the configured geolocation

Optional: ROS integration

The simulator does not need ROS, but the repository also ships a ROS 1 (Noetic) bridge that publishes the vehicle state as topics and renders it in RViz: scripts/sim4cd/ros_viz.py, scripts/sim4cd/sim_ros.py, launch/sim4cd.launch, rviz/basic.rviz, plus the package.xml and CMakeLists.txt that make the repo a catkin package. None of this is used by the Qt GUI or by sim4cd_main.py.

To use it, clone the repo into a catkin workspace and build it there:

mkdir -p ~/catkin_ws/src
cd ~/catkin_ws/src
git clone -b add-python-qt-gui https://github.com/adrianomcr/sim4cd.git
cd ~/catkin_ws
catkin build

MAVROS is also optional, and is what lets PX4 itself appear on the ROS graph:

sudo apt install ros-noetic-mavros
roscd mavros/../../lib/mavros/
sudo ./install_geographiclib_datasets.sh
pip3 install pymavlink

ROS Noetic is Ubuntu 20.04 only, so this whole section is limited to that distro. The Docker image is Ubuntu 24.04 and does not contain ROS.

Legacy Tk GUI

The original Tk GUI is still available in scripts/gui/. It is superseded by the Qt GUI, but it works and uses the same JSON parameter files. See scripts/gui/README.md.