Cooperative Multi-Agent Jamming of Multiple Rogue Drones Using Reinforcement Learning
Panayiota Valianti, Kleanthis Malialis, Panayiotis Kolios, Georgios Ellinas
Abstract
The wide adoption and use of unmanned aerial vehicles (UAVs) has created not only opportunities but also threats to the security of sensitive areas. Thus, effective and efficient counter-drone systems are required to protect these areas. This work tackles this issue by developing cooperative multi-agent jamming techniques using reinforcement learning (RL) to counter the operation of one or multiple rogue drones flying over a sensitive area. The aim of the proposed RL approach is to optimize the joint mobility and power control actions of the pursuer UAVs in order to maximize the received jamming power at the rogue drones aiming at disrupting communication links and sensing circuitry, while at the same time keeping the interference to surrounding pursuer agents below a predefined threshold. The effectiveness of the proposed approach in terms of scalability, learning speed, and agents' final joint performance is demonstrated through extensive simulation experiments for various agent and target configurations.