Deep Reinforcement Learning Based Spinal Code Transmission Strategy in Long Distance FSO Communication

We consider a deep reinforcement learning based Spinal code transmission strategy to reduce resource consumption and improve channel utilization while guaranteeing communication quality in long-distance free space optical (FSO) communication. First, a deep Q network is established to model the channel state-action value function, and then the neural network approximation value function is trained so as to determine the number of Spinal code symbols that should be transmitted for effective communication under current channel conditions. The final simulation results show that compared with the basic spinal code transmission mechanism and the adjustment algorithm based on linear filtering, the average throughput of the system using the proposed algorithm is improved by 26% -34%.