Hierarchical Reinforcement Learning for UAS Swarms

What was done: Our swarm team deployed a swarm using control laws dertermined from model-free hierarchical reinfordement learning (HRL) methodologies designed for formation control and ground-target tracking onto a swarm of 8 custom built UAVs.

My contributions

  1. Designed and wrote the sofware implementation of the swarm controller in ROS (C++ and Python): link to github

  2. Designed and wrote the ground control station frontend as a web app: link to github, oversaw the desgin of the backend with development contribution: link to github

  3. Led field experiments to exercise swarm system and headed the final demonstration field campain a the Joint Interagencey Field Experimentation (JIFX 23-3)

  4. Constructed and configured the custom built fleet of quadrotors, made repairs as needed

  5. Designed and protyped the “swarm release rack” (see video below)

Hardware Used

  • Cube Pilot: Cube Blue Flight Controller

  • Raspberry Pi

  • RF Desgin: RFD900

  • DJI F450

Software/Frameworks Used

  • ArduPilot

  • ROS/MAVROS

  • MAVLink

  • MATLAB

IMG_0018 (1).JPG
IMG_8210.jpg
IMG_8285.jpg