Research On Collaborative Computation Offloading Strategy In Edge Computing

With the development of mobile networks and wireless communication technology,the Internet of Things has greatly developed.The amount of data generated by mobile smart terminals has shown an exponential growth,which has led to higher requirements for data storage,transmission,and processing capabilities.Therefore,existing terminal devices have insufficient storage,low processing power,and limited resources when encountered with massive application services.It is difficult to meet the requirements of low latency and high computing power for tasks,which greatly affect the user experience.For terminal devices with limited resources,computing offloading technology is used to offload all or part of tasks to other computing entities to reduce the load,cost,and energy consumption of local terminals.However,most of the existing computing offloading researches are aimed at cloud-edge-device longitudinal research.Ignoring the computing power of some terminal devices and the horizontal collaboration between multiple edge servers that can greatly improve the performance of computing offloading.In order to solve these problems,this paper proposes a multi-objective overall computing offloading model and a manyobjective partial computing offloading model,and designs efficient multiobjective and many-objective optimization algorithms for the model.The main contributions of this paper are:(1)In order to give full play to the performance advantages of collaborative computing offloading between smart terminal devices and edge servers,a multiobjective collaborative computing offloading model composed of edge servers and terminal devices is designed.The model optimizes the two objectives of offloading latency and device load balancing in the way of overall task offloading.Combined with the characteristics of the model,a multi-objective evolutionary algorithm with load constraints is proposed.The optimal computing offloading strategy is obtained by calculating the Pareto optimal solution for the population.Compared with multiple multi-objective optimization algorithms,it is verified that the algorithm designed in this paper can solve the model faster.(2)Considering the diverse offloading problems,flexible offloading methods,and complex task attributes in the actual edge computing scenario,this paper designs a many-objective collaborative computation offloading model in the multi-edge server scenario.Through the three-level collaborative partial offloading method,offloading tasks from resource-constrained terminals to other computing entities is proposed to solve the above problems.Optimizing the four offloading objectives of subtask offloading latency,parent task execution time,energy consumption,and improving device resource utilization to improve offloading efficiency.Combined with the characteristics of the model,the global evaluation strategy based on angle and distance is proposed,which solves the problem that the relationship between individual selection and Pareto advantage is too local due to the conflict of objectives.In the simulation experiments,the effectiveness of the many-objective collaborative computation offloading model was verified.The Taguchi orthogonal experiments on the design parameter evaluated the influence of the parameter value on the value of each objective function.Based on the Wilcoxon rank sum statistical tests,the algorithm performance was verified in the benchmark test problems.Finally,in terms of the model,the experimental results showed that the performance of the four optimization objectives of the offloaded model was improved by 18%,31%,20%,and 42%,respectively.(3)In the process of solving the above many-objective optimization model,the multi-objective evolutionary algorithm gradually loses the selection pressure due to the increase of the dimension in the objective space,which makes it difficult to guarantee the convergence and diversity of the solution set.In order to solve this problem,this paper proposes an evolutionary algorithm based on information entropy selection strategy.The algorithm calculates the distribution of the current individual in the population based on the neighborhood information around the individual,and at the same time maps its distribution information to a Gaussian function.Evaluating the distribution of individuals in the population by calculating the information entropy of the population.Selecting individuals to enter the next generation population with good distribution performance.The experimental results showed that the algorithm had significant advantages on the standard multi-objective test sets DTLZ and the many-objective test sets Ma F,and it also had better performance in solving the many-objective collaborative computation offloading model proposed in this paper.