Research On Computation Task Offloading Methods Of Mobile Edge Computing For Industrial Internet

With the rapid advancement of the new generation of industrial revolution,higher requirements for multi-dimensional indicators of industrial applications are put forward by the Industrial Internet,such as,reliability,time delay,energy consumption,and quality of user experience.As a key technology of the Industrial Internet,mobile edge computing(MEC)can effectively reduce the time delay for transmitting the data to cloud center,save the energy consumption of mobile devices(MDs),and enhance the computing power of MDs by transmitting the computation-intensive or time-sensitive tasks to the edge serves which are approximate to MDs.Currently,the main goals for most works of computation task offloading in MEC are the collaboration of computing and communication resources and the tradeoff between the system energy consumption and processing delay.However,due to the complex external experiement of Industrial Internet,any changes on different factors will have significant impacts on the offloading decisions,such as the reliability of MDs,the operation time of the MDs,the type of fine-grained computation task offloading,and the complex dependencies between computation tasks.Hence,in the industrial Internet environment,how to determine the offloading decisions of computation tasks to meet the requirements of as many network indicators as possible and be robustly applicable to different application scenarios has greater theoretical value and practical significance.Aiming at the features of the Industrial Internet,such as high reliability,low energy consumption,and low time delay,focusing on the reliability of MDs,fine-grained computation task offloading with long-term operation of MDs,and complex dependencies between computation tasks,the thesis studied the computation offloading methods for MEC in the Industrial Internet.The main contributions and novelties can be summarized as following.1)Aiming at the high reliability of the Industrial Internet,the thesis studied the computation task offloading method with reliability of MDs in MEC of the Indsutrial Internet,and proposed a Heuristic Algorithm based on the Greedy Policy,namely HAGP.In more details,considering the computation offloading decisions will increase the overall system overhead and reduce the economic efficiency of the actual production process when MDs are unreliable,we explicitly defined the reliability of MDs with the residual energy after completing a computation task,and regarded it as an important constraint for determining the offloading decisions.Further,by analyzing the models of energy consumption and time delay for different computing modes,we constructed a mixed integer non-linearly programming problem with the goal of minimizing the overall system overhead which is defined as the weighted sum of energy consumption and time delay.Subseqently,to effectively solve the optimization problem,combining the alternative optimization and greedy policy,we designed and proposed HAGP,by which the optimal offloading decisions and the corresponding optimal CPU cycle frequency or the optimal transmission power can be obtained.Finally,we qualitatively compared the HAGP with 3 algorithms that considers transmission reliability,task reliability,and computing mode reliability.Meanwhile,we conducted extensive simulation experiments to quantitatively compare the HAGP with 3 typical baseline algorithms and the algorithm without considering the reliability of MDs.The results showed that HAGP can be effectively applied to the computation task offloading for multi-user in MEC with considering the reliability of MDs.2)Aiming at the low energy consumption of the Industrial Internet,the thesis investigated the partial computation task offloading for MDs enabled by harvested energy,and proposed a novel algorithm,that is the Lyapunov Optimization-based Partial Computation Offloading(LOPCO).Specifically,considering that the integration of energy harvesting technology and fine-grained partial computation task offloading will bring some challenges for offloading schemes in MEC,we built models for different computing modes and energy harvesting process,and formulated the problem as a non-convex nonlinear optimiazaion problem which aims at minimizing the long-term average energy consumption of all MDs and the edge server.Then,in the non-convex non linearly optimization problem,considering the energy casual constraint and the coupling constraint between offloading decision and resource allocation,we designed and proposed LOPCO based on the Lyapunov optimization and the variable substitution methods with jointly obtaining the optimal offloading ratio,the optimal CPU frequency,and the optimal transmission power.Finally,by comparing the simulation results of LOPCO with a classical baseline algorithm and two state-of-the-art algorithms,we get that the performance of LOPCO is significantly better than others in partial computation offloading for MDs enabled by harvested energy.3)Aiming at the low latency of the Indsutrial Internet,the thesis presented the jointly scheduling and offloading method for dependency-aware computation tasks,and proposed a new algorithm for jointly scheduling and offloading computation tasks based on the highest responding ratio and genetic algorithm,namely HRRO-GA.Concretely,considering that the intelligence and automation will complicate the dependencies between computaion tasks of the industrial applications,while different scheduling and offloading schemes will greatly impact the completion time of the applications,we built mathmatical models for scheduling and offloading processes in ultra-dense MEC networks,and formulated a non-convex mixed integer nonlinear programming problem with minimizing the overall time delay of applications.Then,to solve the problem,we firstly designed and proposed a new algorithm based on the highest responding ratio first for computation offloading in scenario of single MD and single edge server,that is HRRO.Moreover,considering the computing complexity of current algorithms are higher,we extended the HRRO into the ultra-dense MEC system,and proposed a novel low computation complexity algorithm based on the Genetic Algorithm,namely HRRO-GA.Lastly,we made a large amount of simulation experiments,analyzed the impact of different factors,and compared the results with several existing algorithms,it can be seen that HRRO-GA is more suitable for jointly scheduling and offloading the computaion tasks with dependencies in complex MEC system.