Vision-Based Swarms in the Presence of Occlusions

This repository refactors the following article in Python:

F. Schilling, E. Soria, and D. Floreano, "On the Scalability of Vision-based Drone Swarms in the Presence of Occlusions," IEEE Access, vol. 10, pp. 1-14, 2022. [IEEE Xplore] [Citation]

And the following video gives the explanation of the article.

Installation

git clone git@github.com:duynamrcv/vision_flocking.git

Run simulation

The current version anables some type of neighbor selection methods, to change the selection type or other configuration parameters, edit config.py file.

• MODE:
  • metric: robots in predefined distance are choose as neighbors.
  • vision: robots that vision distance and do not occluded are chose as neighbors.
• USE_VORONOI:
  • True: use voronoi to refine neigbor set
  • False: not use voronoi

To run the simulation, please run:

python3 main.py

The data will be saved in *.txt file. To view animation, please run:

python3 animation.py

Note that the parameter export=True is use to export video file, if not, set export=False.

Results

Vision-based neighbor selection	Vision-based Voronoi neighbor selection

Comparison

Paths

Vision-based neighbor selection	Vision-based Voronoi neighbor selection

Scalability

Noise