ADIGE 360 — Autonomous Drone Navigation In GPS-denied Environments

The Adige is the second-longest river in Italy, flowing into the Adriatic Sea. Water always finds its way. So should a drone.

When a micro-drone loses its communication link in a GPS-denied indoor environment, it must be able to navigate autonomously toward open space — preserving onboard sensor data that would otherwise be lost with the vehicle. Like water finding its path, the drone follows the open space.

Code for the paper:

A 360-Degree Vision Dataset for Learning Yaw Control on GPS-Denied Micro-UAVs in Disaster-Response-Relevant Environments Niklas Voigt, Hartmut Surmann — Westphalian University of Applied Sciences (WHS) Accepted at the 2026 IEEE/ASME International Conference on Advanced Intelligent Mechatronics (AIM 2026), Genova, Italy.

📦 Dataset: huggingface.co/datasets/RoblabWhGe/adige-360

# clone repo
git clone https://github.com/RoblabWhGe/adige-360.git
cd adige-360

# virtual env with uv
uv sync
source .venv/bin/activate
cd src

# single training run
python train.py 1337
# or bulk training
./bulk_train.sh

If you use this code or the dataset in your research, please cite:

@inproceedings{voigt_dataset_2026,
  author      = {Voigt, Niklas and Surmann, Hartmut},
  title       = {A 360-Degree Vision Dataset for Learning Yaw Control on {GPS}-Denied Micro-{UAVs} in Disaster-Response-Relevant Environments},
  booktitle   = {2026 {IEEE/ASME} International Conference on Advanced Intelligent Mechatronics ({AIM})},
  eventdate   = {2026-07-07/2026-07-11},
  venue       = {Genova, Italy},
  publisher   = {IEEE},
  date        = {2026},
  pages       = {1--8},
  doi         = {10.1109/AIM65483.2026.11658039},
  isbn        = {979-8-3195-3611-2},
  issn        = {2159-6255},
  eprint      = {2608.19866},
  eprinttype  = {arxiv},
  eprintclass = {cs.CV},
}

• Code in this repository is released under the MIT License.
• Dataset (images and annotations) is released under CC BY-NC-SA 4.0 and hosted on Hugging Face.