Research On File Transfer Optimization For Navigation Domain

When the navigation system transmits files on the private line network,the transmission efficiency of the files will affect the execution of the satellite orbit determination task.Data is transmitted on a network with low bandwidth,and the excess network traffic will make network congestion and increase the latency.In the high-latency scenario,the throughput of the network will decrease,and the transmission of files will have the problem of not reaching the designated terminal on time.Improving the file transmission efficiency on the private line network and making the files arrive the designated terminal on time are the goals to be studied and achieved in this thesis.This thesis takes the file transmission of a system in the field of navigation as the research background.Aiming at the problem of low link throughput in low bandwidth and highlatency scenarios,the reinforcement learning-based congestion control algorithms are introduced to improve the bandwidth utilization and fair allocation of the private line network.According to the number,size and priority of files transmitted by the system,file transmission is divided into different scenarios,optimization algorithms for different scenarios are designed and implemented,and the effect of file transfer optimization is tested and verified.The main contents and achievements of this paper include the following points.(1)Aiming at the problem that files cannot be transmitted within the specified time on the private line network,we set up comparison experiments to analyze the shortcomings of existing congestion control algorithms from the test results of bandwidth utilization and allocation fairness,introduce the TCP＿PPO2 reinforcement learning congestion control algorithm with higher bandwidth utilization and MARLCC reinforcement learning congestion control algorithm with better fairness.According to the file information of the issued command,the file transmission is divided into different scenarios according to the number,size and priority of the files,and different scenarios are analyzed.(2)Using reinforcement learning-based congestion control algorithms instead of the default congestion control algorithm to improve network throughput,using compression algorithm to compress large files,designing optimization algorithms for different file transfer scenarios and implementing adaptive selection of algorithms and dynamic allocation of bandwidth.The TCP＿PPO2 algorithm and the MARLCC algorithm are deployed in the operating system by adding modules;using Mininet simulation software,the performance of the reinforcement learning congestion control algorithms are compared with traditional congestion control algorithms.The experimental results show that the bandwidth utilization of the TCP＿PPO2 algorithm is improved by 14%～32% compared with traditional congestion control algorithms,and the fairness index of the MARLCC algorithm is improved by 6%～42% compared with traditional congestion control algorithms,the superiority of reinforcement learning congestion control algorithms are verified;the performance test of different compression algorithms is carried out,and the zl4 algorithm is selected to be applied to the compression of big files;the file transmission is divided into different scenarios according to count,size and priority,the optimization algorithm of file transmission corresponding is designed.The optimization algorithm realizes the adaptive selection of the congestion control algorithm by the system according to the number of files,design different bandwidth allocation formulas according to the size and priority of the file,realize the dynamic allocation of bandwidth in different file transfer scenarios.(3)Testing the file transfer function of opening and closing the optimization algorithm to verify the effect of file transfer optimization.Using 620 MB data acquisition file,20 MB antenna file and 30 MB related file as test files,designs test cases and conduct file transfer tests,with the file transfer completion time FTCT(File Transfer Completion Time)as the evaluation index.The test results show that after the file transfer optimization algorithm is enabled,the average FTCT of different transfer scenarios is shortened by 20% to 30%,which verifies the effectiveness of the file transfer optimization algorithm implemented in this thesis.