Computation Offloading And Resource Allocation In Mobile Edge Computing Networks

With the rapid development of mobile internet,internet of things and artificial intelligence technologies and the exponential increase of smart mobile devices(SMDs),some new types of resource-intensive mobile applications with low latency and high energy consumption come into being,which leads to the result that the contradiction between the resource-intensive applications and the limited capability of SMDs has been prominent recently.Mobile edge computing(MEC)deploys the cloud servers at the edge of the mobile networks,in proximity to SMDs.Owing to the nearby computation offloading and task execution,MEC has been envisioned as a promising technology to improve the response time of computation service as well as alleviate the computation and energy consumption of SMDs.However,considering the co-channel interference caused by spectrum reuse in heterogeneous MEC networks,the delay of service response from massive computation requests in a centralized MEC network and the high dynamic movement of computation services in unmanned aerial vehicles(UAVs)assisted MEC networks,serious challenges are posed to the realization of MEC and the improvement of the overall system performance.Aiming at the improvement of system performance,such as energy consumption,latency,the tradeoff between energy consumption and latency as well as the computational energy efficiency,this thesis conducts numerous research on the joint computation offloading and resource allocation in the infrastructure based and UAVs assisted MEC networks,subject to the limited resources of computation,communication and storage,the latency sensitive computation tasks and the guarantee of quality of services.The main work and contributions are summarized as follows.Considering the tradeoff between energy consumption and latency in heterogeneous MEC networks,we propose an energy-aware computation offloading and resource allocation scheme,which can be applied to the heterogeneous networks with high density small base stations.Firstly,we introduce the residual energy of SMDs into the definition of the weighting factors of the energy consumption and latency,so as to sense the current service status of SMDs and achieve the adaptive adjustment of the system optimization objective.Under the constraints of the limited energy of SMDs and the latency sensitivity of computation tasks,we formulate the problem of computation offloading,computation and communication resource allocation to minimize the weighted energy consumption and latency.Secondly,due to the coexistence of channel interference and the coupling between computation offloading,channel allocation and transmit power in heterogeneous MEC network,the tradeoff problem between energy consumption and latency is non-convex mixed integer nonlinear programming(MINLP)problem.In terms of the non-convex problem,we decompose the original problem into three subproblems,i.e.,the local computing overhead optimization,channel allocation based on the maximal effective interference and the optimization of power allocation and computation offloading,and propose an iterative search algorithm to find the suboptimal solutions.Finally,numerical results show that the energy-aware weighting factor is of great significance to maintain the lifetime of SMDs compared with the given weighting factor,and the proposed algorithm can obtain lower total overhead than the baseline algorithms.Considering the latency sensitivity of computation tasks in MEC networks with caching,a joint optimization of network resources of computation,communication and storage is considered,and the modified branch-and-bound method and generalized benders decomposition method for computation offloading and resource allocation are proposed.Firstly,in a centralized MEC network,considering the limited resources of computation,storage and spectrum,we formulate the joint optimization of computation offloading,edge caching and resource allocation to minimize the total latency of system with the guarantee of the quality of service.With the coupling of integer variables and continuous variables,the formulated problem is non-convex,which is intractable to solve.Then,based on the constrained relationship of offloading decision and caching decision,we propose the branch-and-bound method with an asymmetric search tree to solve the problem,which has an exponential computational complexity.Finally,in order to lower the computational complexity,a generalized benders decomposition method with integer calibration is proposed to obtain the integer and continuous solutions in iterative manner,and it has a polynomial complexity.Simulation results show that the proposed algorithms can obtain lower latency consumption compared with the baseline schemes,and the performance gap between the branch-and-bound algorithm and the generalized benders decomposition algorithm is small.In terms of the computation cost,the generalized benders decomposition algorithm has less running time.Considering the limited energy of UAV assisted MEC network and the dynamic mobility of the UAV,we model the computation tasks arriving at SMDs as the random task queue and propose a joint optimization model for stochastic computation offloading,resource allocation and trajectory scheduling.Firstly,the task queue update process is established by the combination of local computing bits,offloading bits and edge computing bits,and the energy consumption of the system is formulated.Considering the limited computation resource of SMDs and the UAV,the constrained relationship that the computed tasks cannot be more than the arriving tasks and the trajectory scheduling of the UAV,the average energy consumption minimization problem is modeled.Secondly,in order to find the optimal solutions of time average problem,we adopt the Lyapunov method to analyze the arriving and offloading task queue.The original problem is transformed as the tradeoff problem between the system utility and queue stability.Due to the tradeoff problem is non-convex,it is decomposed into different manageable subproblems.Finally,according to the characteristics of subproblems,we propose an iterative online algorithm which combines alternative direction method of multipliers(ADMM),interior point method and CVX solver to find the optimal solutions.Simulation results illustrate that the proposed algorithm can effectively achieve the balance between system energy consumption and the queue stability compared with the baseline schemes.Besides,the optimal parameter selection of the control coefficient to gauge the tradeoff between the system utility and queue stability is concluded in the experimental discussion,which can be adaptively utilized according to different requirements.Considering the computational energy efficiency of multi-UAV assisted MEC network,we propose a computation offloading and resource allocation scheme to guarantee the system computational energy efficiency in complex MEC network with multi-UAV and multi-obstacle.First of all,under the condition of ensuring the execution of minimal computation tasks,we formulate the joint optimization of the association between SMDs and UAVs,the limited computation capability and transmit power of SMDs,the limited spectrum resource and the trajectory scheduling of UAVs as the computational energy efficiency maximization problem,which is an intractable non-linear fractional problem.Secondly,the Dinkelbach method is applied to transform the original problem into a tractable parametric programming problem about the computational energy efficiency.Due to the coupling between trajectory scheduling of UAVs and other variables,the problem is still non-convex.Finally,we propose an iterative computational energy efficiency maximization algorithm with double-loop structure to solve the problem.In outer loop,the Dinkelbach method is employed to find and update the optimal computational energy efficiency,while a joint optimization problem for user association,resource allocation and the trajectory scheduling of UAVs is solved in inner loop.Simulation results demonstrate that the proposed algorithm can converge in a number of iterations and obtain higher computational energy efficiency than the baseline schemes while guaranteeing the quality of computation services.Besides,the UAVs can succeed in avoiding obstacles to finish the flying mission.