Drone Swarm Tactics using Reinforcement Learning and Policy Optimization

Midshipman Researcher(s): 1/C Elizabeth Gergal

Adviser(s): Professor Ric Crabbe

Poster #28

The goal of this project is to develop offensive and defensive swarm tactics using reinforcement learning (RL). We are simulating battles between two swarms of military fixed wings drones using both pre-programmed and agent learned tactics. The drone agents are developing intelligent behaviors and tactics as a result of the RL algorithms through which they are trained. Implementing RL into the development of swarm tactics will maintain the US military advantage over our enemies and protect America’s national interests.

Full Size Computer Science #28