Research On Edge Computing Task Offloading And Service Migration Scheme In Internet Of Things

In the past few years,the development of the 5th generation mobile communication(5G)along with the deployment of the Internet of things(Io T)has significantly increased.The computing paradigm has improved in transmission rate,deployment density,spectrum efficiency of communication network,popularity of smart devices and low delay applications,and growth of terminal data.However,these improvements ended up resulting in more prominence in the limitations of centralized computing paradigm,which in return suggests a new perspective for research in decentralized computing paradigm.For example,the research on multi-access edge computing(MEC)is still in the early stage with some scientific problems yet to be solved.We believe it is necessary to consider the joint optimization of computing resources and channeling resources for the characteristics of intensive deployment of user equipment and large amount of resource requests in heterogeneous networks.Furthermore,we formulate scientific and effective edge service migration strategies to improve the quality of service(Qo S).This paper studies the strategy of edge computing task unloading and service migration to solve the problems above.The contributions of this thesis can be summarized as follows:1)We constructed a two-level edge nodes(ENs)relay network model,for heterogeneous IT application scenarios.We also designed an optimal energy consumption algorithm(OECA)for the joint optimization of computing and channeling resources to maximize speed and energy efficiency during computing.First,we modeled the energy efficiency in MEC as a 0-1 knapsack problem.Next,we aimed for minimizing the overall energy consumption of the system,by having the system select the computing mode,and allocate the wireless channel resources in an adaptive manner.Finally,we verified the performance of the algorithm by simulation in Python.The simulation results show that OECA increases the network capacity by 18.3% and reduces the energy consumption by 13.1% from the directed acyclic graph algorithm(DAGA).2)In application of internet of vehicles(Io V),consider the optimization problem of intelligent connected vehicle(ICV)edge service migration at urban 2-D intersection.We designed an edge migration strategy based on reinforcement learning(RL)algorithm.First,we extracted the joint state of ICV and RSU under the edge network to ensure the optimal Qo S problem of the migration process is fitted to the optimal solution of the Markov decision process(MDP).Next,we used the attention mechanism to construct its state value model,action set,and reward function.Then,the state value model is trained by a Q-learning agent to obtain the mapping relationship between its action set and reward function.Finally,we ran comparative experiments;we found that the edge migration strategy based on RL algorithm can improve Qo S by about 5.2% from the greedy algorithm.