Authors:

Connor Bradley Frey

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Unmanned Aerial Vehicles (UAVs) are becoming more widely used around the world for attacks against infrastructure. The use of a network of surveillance cameras allows for a simple and cost-effective method to prevent these attacks by using the cameras to spot the UAVs. Reinforcement learning is a method that can allow agents controlling the cameras to find the optimal policy for detecting UAVs. A way to simulate this is by developing a multi-agent reinforcement learning (MARL) environment. A MARL environment would allow each camera and UAV to be controlled by a separate agent which significantly reduces the size of the local action space for each agent. This reduced action space makes it easier for the agents to make decisions and can speed up the runtime when using parallel processing. The use of agents to simulate the attacker and defender is so that we can find the Nash equilibrium strategy such that when the defender employs this strategy the attacker will only have one combination actions that will not hurt their position. This MARL environment would further support research in reinforcement learning algorithms for counter-UAV surveillance. Keywords: [no keywords provided]

Source:

Purdue University / 2025

Topics:

No topics listed

Co-authors:

Connor Bradley Frey