Research And Design Of Node Mobility Prediction And Computing Offloading

Mobile edge computing is a distributed computing model in the context of Io T that can transfer computation,storage,and processing power from small cloud centers to the edge of the network.In this paper,the architecture and characteristic advantages of mobile edge computing,the proximity and location acquisition are described,several emerging application scenarios are analyzed,and the future development of mobile edge computing and the possible challenges are further discussed.In this paper,the research is carried out in two aspects of node mobility prediction and computational offloading methods as follows.A prediction service framework based on mobile edge computing is designed to predict the endpoint location of mobile node travel destinations using a ridge regression optimized echo state network.The model is trained and tested using publicly available datasets.The advantages of the prediction model are verified by comparing with the coordinates of real trajectory location points.A computational offloading scheme with complete offloading is designed,based on the Q-learning algorithm in reinforcement learning,and the computational latency and energy consumption of the multi-mobile device scenario are weighted and summed to optimize as the system cost,and compared with other benchmarks for analysis,and the simulation results show that the method exhibits better performance with a limited number of devices.For the channel time-varying and task randomness characteristics of multi-device multi-input systems,an attempt is made to make devices self-learning offloading strategies,and a dynamic offloading framework based on depth-deterministic gradients is designed to split computational tasks to process data in parallel,reduce the power consumption and latency of local computation and offloading execution,and introduce greedy algorithms for comparison experiments.The advantages of the approach are verified by comparing single-device and multi-device scenarios.In this paper,an echo network prediction model based on ridge regression optimization is designed for the movement characteristics of nodes,and the comparison with the results of actual trajectories proves the good performance of this model.In addition,for the computational offloading delay and energy consumption optimization,a dynamic offloading method based on Q-learning and deep reinforcement learning is designed to reduce the computational cost of system power consumption and delay.