Research On Routing Optimization Of Data Offloading Based On High Mobility Nodes

Due to the explosive growth of global data traffic,bulk data transmission based on delay-tolerant networks faces challenges such as insufficient network resources and limited coverage,so new network architectures need to be introduced to assist bulk data transmission.With the development of Internet of Vehicles and space-air-ground integrated networks,the problems of network capacity and communication distance have been solved,but the high mobility of nodes in these networks and frequent dynamic changes in network topology lead to low efficiency of bulk data transmission.To solve this problem,this paper investigates the data offloading routing optimization problem based on Internet of Vehicles and space-air-ground integrated networks,and the main work is as follows:(1)A bulk data transmission system based on Internet of Vehicles is proposed and the routing of data offloading and transmission is optimized.In this system dedicated vehicles are used as carriers for data offloading,and the physical movement of the vehicles becomes part of the data transmission.An offloading service model is first constructed based on the Manhattan mobility model,then a new delay model is proposed for the data offload and transmission process,and finally the data transmission allocation problem is transformed into a minimum delay path search problem.A pre-RLGA two-step algorithm is designed to solve this problem,in which a Temporal Convolutional Network(TCN)model is used to predict the traffic volume and construct the delay matrix,and a Genetic Algorithm Based on Reinforcement Learning Mechanism(RLGA)is used to train the model to obtain the optimal routing strategy.The experimental results show that the pre-RLGA proposed in this paper reduces the data transmission delay by 9.2%～21.41% compared with other methods.(2)A bulk data transmission system based on Space-Air-Ground Integrated Networks(SAGIN)is proposed,and the routing of data offloading and transmission is optimized.Firstly,the network states and motion states of nodes in SAGIN are mapped by time-expanded graphs.Then,the bulk data transmission task is partitioned into a series of subtasks,each of which is abstracted as an agent,so that the data transmission process of each subtask is abstracted as a Markov decision process,and the final routing optimization problem is transformed into a multi-intelligence cooperative control problem.To solve this problem,the Time-Expanded Graph Based Multi-agent Deep Deterministic Policy Gradient(TEG-MADDPG)is designed to train the network model in order to obtain the optimal routing policy.The experimental results show that TEG-MADDPG reduces data transmission delay by 23.66%～55.54%,reduces packet drop rate by 7.77%～33.98%,and improves system throughput by 34.39%～152.89% compared with other methods.