Deep Reinforcement Learning-based Resource Allocation for Reliable Multi-UAV Communication Networks

In this paper, we study a joint spectrum allocation and power selection management problem for multi-unmanned aerial vehicle (UAV) communication network in the presence of malicious jammers. As the formulated optimization problem is highly non-convex and the communication environment is complexly dynamic, we model the optimization problem as Markov decision process with reinforcement learning, and then propose an Anti-Jamming Resource Allocation algorithm based on Double Deep Q-Network with Prioritized Experience Replay (AJRA-DDQN-PER), which builds upon the standard DDQN framework. This designed deep reinforcement learning (DRL) approach efficiently learns optimal policies through strategic sampling of critical experiences against jamming attacks, significantly improving convergence speed and solution quality. Numerical simulations demonstrate significantly better performance compared to baseline DRL methods, achieving notable reductions in average transmission delay. These results highlight the algorithm’s potential for practical UAV swarm communication systems and underscore the value of artificial intelligence (AI) techniques in communication networks.