GitHub - AkgunOnur/DroneSwarm

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 49 Commits
.ipynb_checkpoints		.ipynb_checkpoints
assets		assets
models_centralized		models_centralized
models_decentralized		models_decentralized
models_planner		models_planner
sac_discrete		sac_discrete
.gitignore		.gitignore
Grid Space.ipynb		Grid Space.ipynb
README.txt		README.txt
Untitled.ipynb		Untitled.ipynb
agents_positions.pkl		agents_positions.pkl
animate.py		animate.py
config.py		config.py
planner.py		planner.py
point_mass_formation.py		point_mass_formation.py
point_mass_formation_discrete.py		point_mass_formation_discrete.py
quadrotor_dynamics.py		quadrotor_dynamics.py
quadrotor_formation.py		quadrotor_formation.py
quadrotor_formation_no_mp.py		quadrotor_formation_no_mp.py
scanned_percentage.py		scanned_percentage.py
test_discrete.py		test_discrete.py
test_planner.py		test_planner.py
train_discrete.py		train_discrete.py
train_planner.py		train_planner.py
trajectory.py		trajectory.py
waypoints.py		waypoints.py

Repository files navigation

Running GPG requires install deep graph library (https://www.dgl.ai/) and compatible version of PyTorch.


This folder contains code for training GPG as well as reloading parameters for larger swarms.

To run this you must first install gym_formation according to the instructions inside gym_formation


If you wish to train GPG from scratch, run main.py. 
If you wish to watch formations, run parameter_reload.py. parameter_reload.py loads model from ./logs. 

A simple arrowhead formation is specified in constr_formation_flying.py (inside gym_formation/gym_flock/envs/)
To increase the number of agents simply change the parameter self.n_agents in constr_formation_flying.py, line 39.


Lastly, if you wish to use the VPG baseline, uncomment line 23 and comment out line 22 in main.py. 

Thank You Very much