Research On Energy Efficiency Optimization Based On Computation Offloading In Mobile Edge Computin

With the continuous development of information technology and the emergence of computation-intensive applications such as mobile gaming,virtual reality,self-driving vehicles and augmented reality,latency-constrained user applications require higher computing power to meet the growing computational demands.However,the computing power and battery capacity of mobile terminals do not meet this high demand,so the industry has proposed Mobile Edge Computing(MEC)as a technology to cope with it.By using this technology,mobile terminals can offload computationally intensive or delay-constrained computing tasks to edge servers with high computing power,thus saving energy consumption of mobile terminals and improving the efficiency of completing computing tasks.However,the impact of different computation offloading strategies on the energy consumption used by mobile terminals and the completion time of computation tasks is different,and an efficient offloading method can effectively reduce the energy consumption and latency of completing computation tasks.This paper focuses on the efficient offloading strategies designed for different MEC systems to improve the performance of MEC systems and reduce the energy consumption of mobile terminals through such specific offloading strategies.The research in this paper is as follows:(1)Partial Computational Offload Scheme for Delay Constraint.For a three-tier MEC system containing multiple Mobile Devices(MDs),multiple Helper Devices(HDs),and a Base Station(BS),this paper proposes an a priori empirical reference vector guided multi-objective optimization evolutionary algorithm RVEAPE to solve the total energy consumption of MDs and HDs minimum problem.In addition to considering the effects of task division ratio and offloading policy on the offloading performance,the computational and communication resources of MD and HD are also considered,and a constrained mixed integer nonlinear programming problem is modeled accordingly.To solve this problem,this paper first converts the problem into a multi-objective optimization problem and then uses the RVEAPE algorithm to solve the problem.Simulations show that the algorithm significantly outperforms other existing baseline methods in terms of the total energy consumption of MD and HD in MEC network systems.(2)Binary Offloading Scheme of Energy-consumption and Time-consumption Tradeoff.For a MEC system with multiple edge devices and a single base station,this paper designs a deep reinforcement learning-based offloading scheme and a resource allocation scheme DDORS.The problem of minimizing the weighted total cost of delay and energy consumption to complete the task is solved by this DDORS scheme.First,the average cost of all edge devices is formulated as a mixed integer nonlinear programming problem.Then,the original problem is decomposed into two subproblems: the local computing problem and the edge computing problem,and optimization techniques and Deep Reinforcement Learning(DRL)are used to obtain the optimal solution to reduce the weighted total cost of delay and energy consumption to complete the computing task.Simulation results show that DDORS may reduce the cost significantly compared to various baseline schemes.