Research On Edge Computing Task Offloading Problem For Industrial Real-time Computing

In the industrial application scenario of edge computing,the problems of insufficient computing capacity and unbalanced load of edge server are often encountered in the face of constantly changing computing requirements.Task offloading becomes an effective means to solve these problems and can efficiently utilize computing resources.However,due to the problem that the traditional network cannot obtain the network state in real time during the task offloading process,it is not possible to plan the transmission path in line with the real-time requirements and formulate a reasonable strategy for the task offloading.Therefore,a edge computing task offloading framework oriented to industrial real-time computing is proposed in this thesis.The transmission path calculation and strategy formulation of task offloading are studied,and corresponding solutions are proposed to improve the performance.Firstly,combined with the real-time requirements of edge computing task offloading transmission in industrial Internet,the edge network is constructed by utilizing the characteristics of SDN transfer control separation,and a edge computing task offloading framework based on SDN collaboration is proposed.It can unify supervision of computing and network resources and can dynamic and flexible data forwarding and efficient centralized management.Secondly,in order to meet the increasing number of tasks and the real-time requirements of tasks in the industrial internet task offloading scenario,and reduce the transmission delay of task offloading,a task offloading transmission path method is proposed to meet the requirements of delay-sensitive flow and large-bandwidth flow of industrial tasks.The method first build edge computing task flow model,the path transmission calculation problem is transformed into multi-constrained shortest path optimization problem.Then,according to the status of network information,Lagrange relaxation method is used to calculate the transmission path satisfying the task requirements.Thus,it can realize the discharge of industrial network heterogeneous task flow transmission and reduce transmission delay.Finally,a cross-domain request oriented task offloading algorithm is proposed to reduce task offloading delay and realize load balancing among edge servers under different load and computing capacity of edge servers.Based on the time delay and energy consumption requirements,the offloading system model is established,and the cost function of minimizing time delay is set with energy consumption as the penalty term.After that,the traditional artificial bee colony algorithm was improved,which adopted reverse learning to initialize the population and combined multi-dimensional and dynamic population to update the population.Test function was used to verify the effectiveness of the improved algorithm,and the improved algorithm was used to solve the task offloading strategy.The load balance between edge servers can be realized and the offloading efficiency can be improved by using the solution strategy.