Deep Reinforcement Learning-Based Relay Selection Algorithm in Free-Space Optical Cooperative Communications

Relay-aided free-space optical (FSO) communication systems have the ability of mitigating adverse effects link disruption by dividing a long into several short links. In order to solve relay selection (RS) problem in decode and forward (DF) relay-aided FSO system, we model scheme as Markov decision process (MDP). Based on dueling deep Q-network (DQN), DQN-RS algorithm is proposed, which aims at maximizing average capacity. Different from relevant works, switching loss between nodes considered. Thanks advantage cumulative rewards reinforcement learning (DRL), our simulation results demonstrate that proposed outperforms traditional greedy method.