Skip to content

Decentralized policy network for time-optimal multi-drone flight using multi-agent reinforcement learning.

Notifications You must be signed in to change notification settings

KafuuChikai/Dashing-for-the-Golden-Snitch-Multi-Drone-RL

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 

Repository files navigation

Dashing for the Golden Snitch: Multi-Drone RL

0. Introduction

  • A multi-agent environment for time-optimal motion planning. This repository presents a decentralized policy network for time-optimal multi-drone flight using multi-agent reinforcement learning.
  • This project is a reimplementation of gym-pybullet-drones, optimized for multi-agent scenarios. We have adjusted the code to make it more suitable for handling a large number of agents simultaneously.
  • We customize PPO in a centralized training, decentralized execution (CTDE) fashion, based on stable-baselines3 and inspired by the on-policy(MAPPO) repository.

Demonstration Video

Demonstration

Real-world experiments with two quadrotors using the same network achieve a maximum speed of 13.65 m/s and a maximum body rate of 13.4 rad/s in a 5.5 m x 5.5 m x 2.0 m space across various tracks, relying entirely on onboard computation.

Related Papers

The public and release version is coming soon...

About

Decentralized policy network for time-optimal multi-drone flight using multi-agent reinforcement learning.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published