Key Technology Research Of Spinal Code Based On Machine Learning In Long Distance Free Space Optical Communication

In long-distance free-space optical(FSO)communications,the signal strength is attenuated by factors such as turbulence,scattering,and absorption,and atmospheric turbulence is the most important factor.Atmospheric turbulence will interfere with the transmission of the light beam,reduce the performance of the optical communication system,and ultimately affect the communication quality.In addition to adopting traditional error control technology,it is of great significance to seek excellent channel coding and decoding technology and study data transmission mechanisms that can resist long delays and high error codes in order to improve the reliable transmission capacity of free space optical communication.The digital fountain code is a coding method that does not have a fixed coding rate constraint.It can automatically adapt to changes in the communication link without knowing the channel status,and has become a forward error control coding technology that makes full use of the channel transmission capacity under harsh conditions.Spinal code is a flexible,small code length digital fountain code,it can continuously generate an unlimited number of coded symbols,with a truly rateless characteristic.In addition,Spinal codes have very good robustness to interference and noise in the channel,and can still have very good performance even when the channel is poor.Based on the above characteristics,Spinal codes have very good application advantages in long-distance free-space optical communication.However,FSO communication,as a wireless communication mode of information transmission in atmospheric channel,the specific application and key technology of Spinal code still need to be further studied due to the influence of channel environment and other factors.Machine learning method has great significance for solving the specific application of Spinal code in FSO communication due to its strong characteristic representation and autonomous decision-making ability.This paper first studies the basic principle of Spinal codes and the optimized design scheme of Spinal codes based on interleaving technology,then introduces the characteristics of free space optical communication channels,and finally analyzes its application advantages in free space optical communication according to the characteristics of Spinal codes.Aiming at the problem of extremely low feedback retransmission efficiency caused by large path loss,long transmission time,and lack of feedback channel resources in star-to-ground optical communication,a Spinal code compilation based on Q learning algorithm Code strategy.The main goal of this strategy is to reduce the number of retransmissions,through adaptive learning to determine the number of coded symbols that should be sent each time,thereby optimizing decoding latency and improving transmission efficiency.It has been verified by experiments that under the same channel conditions,the average number of retransmissions of the proposed transmission strategy is maintained at about 1,which is about 94%-99% less than the traditional Spinal code and less than the linear Spinal code adjustment method.About 78%-92%.Aiming at the long-distance free-space optical communication scenarios with complex and changeable channels,a Spinal code transmission strategy based on deep reinforcement learning is proposed.This strategy first establishes a deep Q network,and then trains the deep neural network to fit the state-action value function to select the action that can get the most value as the output,thereby obtaining the number of Spinal code encoding symbols that should be transmitted for effective communication in the current channel state.It has been experimentally verified that,compared with the adjustment method based on linear filtering and the traditional Spinal code transmission method,the transmission strategy proposed in this chapter improves the throughput by about 26%-33% under the same channel conditions.