Drone Swarming Tactics Using Reinforcement Learning and Policy Optimization
| Author | Midshipman 1/C Elizabeth K. Gergal, USN (United States Naval Academy) |
|---|---|
| Published | 2021 |
| Format | Research paper |
TL;DR — A 2021 Naval Academy Trident Scholar report that trains simulated fixed-wing drone swarms with reinforcement learning to see what tactics emerge.
Trident Scholar Project Report No. 506 (USNA-1531-2), an honours research project by a senior midshipman at the US Naval Academy, approved for public release. A Naval Academy research poster on the same project names Professor Ric Crabbe as adviser.
The project pits two simulated swarms of fixed-wing military drones against each other, comparing pre-programmed tactics with tactics learned by reinforcement learning and policy-optimisation methods, and reports how the learned behaviour performs. It is a simulation study, not a fielded system.
What it is: an undergraduate honours research report from a service academy, cleared for public release; useful as a snapshot of academic work on swarms, not as a peer-reviewed result. A verified public host was not found (DTIC was offline when checked).
Where this came from
- A copy saved in October 2026.
- Related project poster: USNA midshipman research poster.
59 pages. A PDF copy is held in Kyle’s private archive; it is not linked or shared here.