Research On Load Balancing Routing Strategy For LEO Network Based On Traffic Prediction

With the development and advancement of communication technology,the demand for global coverage communications is booming.Due to the advantages of low-orbit satellite network(LOW Earth Orbit,LEO),which has the advantages of low communication latency,large coverage area,simple access and low launch cost,it can effectively realize the seamless real-time information transmission around the world.However,low-orbit satellite networks also have the disadvantages of high dynamic topology changes,limited storage and computing resources,and uneven distribution of global traffic,which bring great challenges to network routing design.The proposed load balancing routing algorithms mainly have the problem of poor perception of network traffic congestion,resulting in the problem that measures to alleviate congestion are often lagging behind.In addition,it is difficult for existing routing algorithms to balance flexibly adapting to changes in the network environment and avoiding local optimization.Therefore,in view of the above problems,the traffic prediction technology is used to solve the routing problem,and the semi-distributed routing algorithm is innovatively combined with the traffic prediction technology to propose a semi-distributed load balancing routing strategy based on traffic prediction.The main innovation points of this paper are listed as follows:1.Aiming at the problem that the proposed traditional time series model cannot adapt to the self-similarity of traffic,this study intends to design a traffic prediction algorithm based on Empirical Mode Decomposition(EMD)and Attention Mechanism(AM).First,the original traffic sequence was decomposed into multiple independent components of single frequency by EMD method,and predicted by LSTM neural network respectively.Finally,the single layer attention mechanism was used to assign different weights to different hidden layer states by mapping weighting and learning parameter matrix,so as to strengthen the influence of important information.The simulation results show that the traffic prediction algorithm proposed in this paper can adapt to the self-similar characteristic of traffic and effectively improve the prediction accuracy.2.Under the background of high dynamics and uneven traffic distribution of low-orbit satellite network,a new semi-distributed hybrid strategy load balancing algorithm based on traffic prediction is proposed in this paper,aiming at the problems that centralized routing scheme consumes large storage resources,cannot adapt to highly variable network topology,and distributed routing is prone to route loop phenomenon.First,each satellite node determines the congestion state according to the traffic value predicted in the future.Then the two-hop range of node status information perception.At the same time,the adaptive link cost is designed.The node changes the weight of the link cost and calculates the double-hop link cost according to different packet types.After the path direction is determined,the double-hop route calculation starts from the source node,and the intermediate node only forwards,so as to realize semi-distributed route forwarding.The simulation results show that the proposed algorithm can effectively reduce the packet loss rate and improve the system throughput compared with the existing schemes.The end-to-end delay can also be reduced when the terminal sending rate increases,which solves the problem that the centralized load balancing strategy cannot adapt to traffic changes and the distributed load balancing strategy falls into local optimal.