DQN-based decentralized multi-agent JSAP resource allocation for UAV swarm communication

It is essential to maximize capacity while satisfying the transmission time delay of unmanned aerial vehicle (UAV) swarm communication system. In order to address this challenge, a dynamic decentralized optimization mechanism is presented for the realization of joint spectrum and power (JSAP) resource allocation based on deep Q-learning networks (DQNs). Each UAV to UAV (U2U) link is regarded as an agent that is capable of identifying the optimal spectrum and power to communicate with one another. The convolutional neural network, target network, and experience replay are adopted while training. The findings of the simulation indicate that the proposed method has the potential to improve both communication capacity and probability of successful data transmission when compared with random centralized assignment and multichannel access methods.