Computation Offloading Strategy For Energy Consumption Optimization In Mobile Edge Computing

Energy consumption is a key indicator for edge computing to achieve energy-saving,green and carbon-neutral development,which affects the deployment of edge computing and the formulation of resource scheduling decisions.Controlling the energy consumption of edge computing is an urgent demand to accelerate the deep integration of edge computing and 5G,promote the construction of edge computing and achieve edge carbon-neutralization.The thesis comprehensively analyzes the energy consumption factors of each edge computing subject and builds an energy consumption model.This thesis builds a computing offloading model based on 0/1 offloading and partial offloading mode in edge computing.The offloading strategy based on evolutionary algorithm and the task partitioning strategy based on reinforcement learning are designed to solve the energy consumption optimization problem under different offloading modes.The main contributions are as follows:(1)Computation offloading model for energy consumption optimization.The thesis constructs a computing offloading scenario of multi-task and multi-edge node for 5G edge computing,respectively builds computing offloading models for 0/1 offloading and partial offloading,and comprehensively analyzes the energy consumption factors in the offloading process to construct the edge system energy consumption model.(2)Multi-task complete offloading strategy based on evolutionary algorithm.This strategy considers both task offloading location decision and computation resource allocation,and proposes a memory search evolutionary algorithm based on the former evolutionary algorithm.The algorithm improves the effective decision rate in multi-constrained edge computing environment by compressing decision coding and designing new mutation methods.The experimental results show that the memory search evolutionary algorithm can effectively obtain the optimal offloading strategy and reduce the energy consumption of the edge system.(3)Multi-task partial offloading strategy based on reinforcement learning.In this strategy,the task is offloaded with fine granularity partition.By constructing a Markov decision process serialization model based on computation offloading process,the task offloading is characterized as a continuous control problem of fine granularity offloading,and then the offloading strategy based on D3 QN algorithm is designed to solve the problem.The experimental results show that the new algorithm can acquire a better task partition offloading strategy in heterogeneous edge environment,so as to maintain a low level energy consumption.