Geometric Model Predictive Path Integral for Agile UAV Control with Online Collision Avoidance

Pavel Pochobradský, Ondřej Procházka, Robert Pěnička, Vojtěch Vonásek, and Martin Saska

This repository contains the open-source implementation of our paper titled 'Geometric Model Predictive Path Integral for Agile UAV Control with Online Collision Avoidance'.

Paper and Citation

If you use this work in your research to build on it, or for comparison, or if you find it useful, please consider citing the following paper:

@article{pochobradsky2026geometric,
  title={Geometric model predictive path integral for agile uav control with online collision avoidance},
  author={Pochobradsk{\`y}, Pavel and Proch{\'a}zka, Ond{\v{r}}ej and P{\v{e}}ni{\v{c}}ka, Robert and Von{\'a}sek, Vojt{\v{e}}ch and Saska, Martin},
  journal={IEEE Robotics and Automation Letters},
  year={2026},
  publisher={IEEE}
}

Simulation Environment: The simulation experiments presented in the video were conducted using the MRS UAV FlightForge Simulator, a high-fidelity Unreal Engine–based simulator developed by the Multi-Robot Systems (MRS) group at CTU.

Installation and Usage

For installation instructions, please choose the guide that best suits your setup from the installation folder:

Apptainer

Please follow the instructions in the Apptainer installation guide.

Docker

We also provide a Dockerfile to build a Docker image with all the dependencies installed and the simulation ready to run.

• For more details on how to build and run the Docker image, please follow the instructions in the Docker installation guide.

Please write to us if you face any issues.

Contact

For any questions, issues, or help, please either contact Pavel Pochobradský or Ondřej Procházka.