Research On Multiuser Dynamic Task Offloading Strategy For Mobile Edge Computing

Mobile cloud computing(MCC)that is regarded as an emerging computing paradigm enables mobile devices to offload their computation tasks to nearby resource-rich cloudlets to enhance computation capability and reduce energy consumption of mobile devices.Mobile edge computing(MEC),as an important part of 5G technology,has innovated the traditional cloud computing and enriched the application scope of MCC.As a promising computing paradigm,MEC can dramatically promote the computation capability and prolong the lifetime of mobile devices by offloading computation-intensive tasks to edge cloud.However,the existing research lacks the research on the recovery strategy of computing offloading between the tasks-dependency after the failure of wireless communication.At the same time,the research on multi-user dynamic task offloading of MEC with different SWIPTs(Simultaneous Wireless Information and Power Transfer)modes is also less involved.Based on the latest research of MCC and MEC,this thesis mainly studies the following three aspects: the robust computing offloading strategy in MCC,the multiuser computing task offload strategy in MEC with energy harvesting,and the resource scheduling strategy in frequency division(FS)SWIPT system.The main contributions of this thesis are summarized as follows:1.This thesis studies the offloading scheduling strategy in the case of computing offloading failure in cloudlet.Due to the mobility of mobile devices and the admission of cloudlets,the connection between mobile devices and cloudlets may be unstable,leading to inappropriate offloading decision and even offloading failure.Therefore,this dissertation proposes a robust computation offloading strategy with failure recovery(Ro FFR)in an intermittently connected cloudlet system in order to reduce energy consumption and shorten application completion time.First,this thesis proposes an optimal cloudlet selection policy when multiple cloudlets near mobile devices are available.Then,the Ro FFR problem is described as two optimization problems,i.e.,local execution cost minimization problem and offloading execution cost minimization problem while satisfying constraints of the tasks-dependency and application completion deadline.By solving both optimization problems,the thesis designs a distributed Ro FFR algorithm for CPU clock frequency configuration in local execution,transmission power allocation and data rate control in cloudlet execution.Finally,experimental results show that our proposed distributed Ro FFR algorithm outperforms several existing offloading schemes in terms of application completion cost and offloading data rate.2.We study the multiuser computation offloading scheduling in MEC with energy harvesting.In light of the energy-constrained characteristics that mobile devices can obtain energy from the radio frequency(RF)based signals,this dissertation obtains the dynamic offloading and resource scheduling optimization model of a multi-user MECSWIPT system by using a real non-linear energy harvesting model.Meanwhile,a dynamic partial computation offloading strategy is proposed by optimizing mobile devices' clock frequency,transmit power and offloading ratio in order to minimize energy cost of mobile devices.More specifically,the thesis first formulates an energy cost minimization problem under the constraints of task completion time and fixed MEC's capacity.In order to solve non-convex optimization problems,an asymptotically optimal algorithm is then proposed to derive the optimal polices for clock frequency control,transmission power allocation,offloading ratio and power splitting ratio.Finally,the simulation results demonstrate that the proposed algorithm not only converges within a few iterations,but also significantly reduces system energy consumption.3.In this thesis,the scheduling problem of FS in SWIPT on MEC is also studied.First,this dissertation proposes a multi-user frequency division multiplexing scheme for SWIPT in MEC offloading and formulate a mixed integer nonlinear programming model(MINLP)for minimizing energy consumption.Then,based on the analysis of the concave-convex characteristics of the optimization model,the problem is transformed into a convex optimization problem by using the theory of variable relaxation and nonconvex optimization.Subsequently,two kinds of optimization algorithms are designed:i)the iterative optimization algorithm based on Lagrange dual method to relax optimization problem,and ii)the integer programming method based on branch and bound where the previous iterative algorithm is used as its basic algorithm for each step of calculation,and a global optimal algorithm is designed for allocating transmission power,offloading strategy,dynamically adjusting of local computing ability and selecting strategy of receiving energy channel.Finally,the simulation results verify the efficiency of proposed scheduling strategy of FS in SWIPT and show that it has better performance of energy minimization in MEC offloading compared with existing algorithms.