Multi-UAV Cooperative Tasking for Attack Detection and Coverage Using Multi-Task Reinforcement Learning
Jin Yu, Chen Lü, Ya Zhang · 2025
This paper presents an innovative end-to-end multi-task reinforcement learning framework designed to enhance the security and operational stability of unmanned aerial vehicles (UAVs) under complex communication attacks. Focusing on multi-UAV cooperative coverage scenarios, the framework enables rapid response and recovery of compromised data by classifying attack types and predicting UAV behavior. First, a multi-task hierarchical attention model is employed to capture the temporal dependencies in communication data, thereby improving the recognition of hybrid attack patterns. This model integrates attack detection and recovery processes. Second, an adaptive pheromone coverage system is introduced to facilitate multi-UAV collaboration with high scalability. Finally, the entire decision-making process is formulated as a multi-task Markov decision process, optimizing the decision strategy to improve overall system performance. Experimental results demonstrate that the proposed framework effectively detects hybrid attacks and predicts UAV actions, ensuring the security and stability of multi-UAV cooperative coverage.