Research On Congestion Control Algorithm Based On Network Congestion Prediction In LEO Satellite Network

With the rapid development of communication technology,satellite communication is becoming more and more mature,low earth orbit(LEO)satellite networks are gradually integrated into the next-generation network infrastructure,which can provide a variety of services in the fields of mobile communications,navigation and meteorology.However,the instability of the link in the low-orbit satellite network and the frequent network topology switching lead to the high dynamic of the network link,which make it difficult for the existing TCP congestion control mechanism to provide stable high throughput performance and high adaptive transmission performance.In this paper,the following two congestion control algorithms are proposed for the non-stability and high dynamics of low-orbit satellite network links,hoping to solve the shortcomings of TCP congestion control mechanism,so as to effectively improve the transmission efficiency of the network.The high bit error rate of LEO satellite link mainly comes from the influence of routing failure,rain attenuation,shadow and multipath interference.In order to solve this problem,this paper proposes a congestion control algorithm called RCC-TCP,which mainly includes the following two contributions:First,it puts forward a network congestion prediction algorithm based on long Short-Term Memory(LSTM),which can predict the degree of network congestion by analyzing the characteristic information on the data message timing.Secondly,the TCP congestion control strategy based on the degree of network congestion is proposed,which establishes the mapping relationship between the degree of network congestion and the growth rate of congestion window through the logics function curve,which can effectively avoid the problem of the decrease of transmission rate caused by non-congestion packet loss.Finally,it is verified by simulation that RCC-TCP has a 10.4%-16.3%throughput performance improvement compared to TCP CUBIC under the conditions of drop rate of 0.001%,1%and long delay of 80ms-200ms.The periodic change of LEO satellite network topology leads to the dynamic change of TCP packet transmission delay,which further interferes with TCP protocol's judgment of network congestion,and makes it difficult for TCP transmission rate to match the continuous switching link state.In view of this problem,this paper proposes a congestion control algorithm called TCP-BQLP,which mainly includes the following three contributions:1)According to the network state parameters,combined with LEO satellite network conditions,the bottleneck link queue length estimation algorithm is proposed on the basis of the algorithm.2)A queue length prediction model based on the ARIMA model is proposed,which can overcome the non-stable and nonlinear characteristics of link estimation data and has high prediction accuracy.3)A congestion window dynamic generation algorithm based on the beta distribution cumulative function curve is proposed,which can generate a congestion window function that conforms to the trend of link bandwidth fluctuation in a certain time slice.Through simulation verification,TCP-BQLP has both better latency performance than TCP Vegas and better throughput performance than TCP CUBIC.TCP-BQLP has the ability to improve the match between transmission rates and network bandwidth.