The Research On Model And Algorithm Of Energy Consumption Optimization In Mobile Edge Scenarios

With the mobile applications develop rapidly,mobile edge computing(MEC)has been widely studied in different fields,which provides a variety of cloud resources(such as computing resources and storage resources)in closer proximity to mobile devices,reducing the completion latency of tasks and the energy consumption of local mobile devices by offloading intensive computing tasks to nearby servers.However,edge servers deployed in static scenarios have the limited coverage,in addition,the energy consumption and latency balance issues of the MEC system were rarely considered in previous offloads,which may increase the total cost of the mobile edge system.Therefore,this paper proposes vehicle-mounted mobile edge mechanism.(1)Describing the joint offloading decision and resource allocation problem of vehicle-mounted edge computing system as a Mixed Integer Nonlinear Programming(MINLP).At the same time,the energy consumption model and delay model of the system processing task are constructed on the basis of considering the energy consumption and running time targets of the vehicle-mounted mobile edge system.In addition,taking into account the balance of energy consumption and delay of the system,a penalty function is added to the defined MINLP in particular.(2)Aiming at the phenomenon that particle swarm optimization algorithm is prone to premature convergence,so a new update strategy is adopted for the two important parameters of inertia weight and learning factor,respectively,and designs an improved particle swarm optimization algorithm(IPSO),and the performance of IPSO algorithm is proved by simulation experiments.(3)The paper designs the specific structure of deep neural network(DNN),adopts the architecture of two hidden layers,and designs the loss function in the training process as a combination of cross entropy and mean square error.In addition,combining the advantages of deep neural network(DNN)and IPSO,the Deep Particle Swarm Optimization Offloading Framework(DPO)is proposed.In order to meet the needs of different situations,this paper designs three different working states of DPO,so as to meet the task requirements and infer the optimal uninstall policy.The performances of DPO,IPSO,Artificial Fish Swarm Algorithm(AFSA),Full Local Execution(FLE),and Greedy are compared and analyzed through the simulation experiments.The simulation results show that the DPO framework proposed in this paper is better than the other four offloading schemes in terms of energy consumption and runtime.