Multi-UAV Telecommunication Inspection against Intelligent Jamming Attacks: A Reinforcement Learning Framework

This paper designs a scenario for unmanned aerial vehicles (UAVs) to assist in telecommunication tower inspection. There are opportunities for further exploration in several areas: 1) Comprehensive practical constraints should be considered; 2) Strong or intelligent attacks should be considered; 3) Single-UAV scenarios should not be the only scenarios considered. This paper focuses on a task where a swarm of UAVs fly within a certain space to inspect the signal strength of telecommunication towers, while also introducing intelligent interference attacks. Specifically, the UAV swarm needs to use a path planning algorithm with excellent robustness to counter this intelligent jamming attack, complete inspection from multiple telecommunication towers, and reach the destination within a specified time limit, while satisfying various practical constraints, including kinematic constraints, no-fly zone constraints, and task deadline constraints. This paper adopts the Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm from Multi-Agent Reinforcement Learning (MARL) to solve this problem. This work gradually illustrate the effectiveness and robustness of defense measures in scenarios with no jamming attacks, fixed jamming attacks, and intelligent jamming attacks.