Research On Resource Management Problems In Mobile Edge Computing

The recent advancements in Internet of Things(IoT)and wireless communication technologies have paved a path towards the explosive growth of mobile devices,data traffic,new computation-intensive and delay-sensitive applications.However,due to heavy backhaul load and long service delay,the traditional centralized cloud computing cannot support such applications with stringent low-latency,high bandwidth,and localization processing requirements.Mobile Edge Computing(MEC)is a promising paradigm which deploys considerable computation and caching resources at the network edge.Exploiting MEC,mobile users can have proximity real-time computing and localization processing capabilities which can reduce network latency and enhance user experience.Mobile users in MEC can offload their compute-intensive and latency-sensitive applications from local servers to edge servers,to alleviate their computation limitations and prolong their battery lifetime.Moreover,edge servers can pre-cache some user requested files or popular content for a short content delivery delay and high user experience.However,without taking computing and caching resources into account,current wireless network resource allocation and deployment scheme cannot be applied in MEC networks.To this end,this dissertation mainly focus on how to integrate computation and caching with wireless communication.This dissertation utilizes heterogeneous networks and vehicular networks as the typical scenarios,to study the problems of joint computation offloading and user association,joint computation offloading and resource allocation,joint computation offloading and load balancing,and edge caching.Specifically,the main contents and contributions of this dissertation are summarized as follows.(1)Joint computation offloading and user association in heterogeneous networksThe finite battery energy and resource-constrained mobile devices,dense deployment of MEC servers,and complexity task dependency relationship pose significant challenges for supporting the computation-intensive and delay-sensitive services.To address the above challenges,this dissertation first proposes a novel two-tier computation offloading framework in heterogeneous networks based on task dependency relationship.Then,with taking task dependency relationship,this dissertation formulates joint computation offloading and user association problem as a mixed-integer non-linear programming problem to minimize overall energy consumption.Based on problem decomposition and semi-definite program,this dissertation develops an efficient computation offloading algorithm by jointly optimizing user association and computation offloading where computation resource allocation and transmission power allocation are also considered.Numerical results illustrate that the fast convergence of the proposed algorithm,and demonstrate the superior performance of our proposed algorithm compared to state of the art solutions.(2)Deep reinforcement learning empowered offloading in heterogeneous networksThe coexistence of macro-cell base stations and small base stations,heterogeneous computation capabilities of devices and edge servers,and different application requirements make it difficult to design an efficient computation offloading scheme.To address this problem,dissertation first proposes an cloud-edge-end orchestrated wireless network,where all devices and base stations have computation capabilities and can concurrently support local computing and computation offloading.Then,dissertation formulates the joint computation offloading and resource allocation problem to minimize system energy consumption as the form of deep reinforcement learning.Finally,this dissertation proposes a new deep reinforcement learning algorithm by optimizing computation offloading and resource allocation.Numerical results demonstrate that our proposed algorithm significantly outperforms the benchmark policies in terms of system energy consumption.Extensive simulations show that the learning rate,the discount factor,and the number of devices have influence on the performance of the proposed algorithm.(3)Joint offloading and load balancing in vehicular edge computing and networksTo alleviate the pressure of resource constrained vehicles and improve resource utility of road-side units,this dissertation proposes a joint computation offloading and load balancing problem in vehicular edge computing and networks.First,this dissertation incorporates the transmission time based on IEEE 802.11 p protocol and the computation time to derive the task processing delay.Then,this dissertation formulates the problem as a system utility maximization problem,and develop a low-complexity algorithm to jointly optimize selection decision,computation resource allocation and offloading ratio.Numerical results illustrate that the proposed algorithm exhibits fast convergence,and demonstrate the superior performance of our proposed algorithm compared to state of the art solutions.(4)An efficient and secure edge caching in vehicular edge computing and networksDynamic network topology,time-varying wireless channel state,and weak trust among vehicles make it difficult to design an efficient and secure edge caching in vehicle edge computing networks.To this end,this dissertation first integrates deep reinforcement learning and permissioned blockchain into vehicular networks.This dissertation proposes a blockchain empowered distributed edge caching framework where vehicles perform edge caching and base stations maintain permissioned blockchain.Then,this dissertation exploits the advanced deep reinforcement learning to design an optimal edge caching scheme for maximizing system utility.Finally,this dissertation introduces a new block verifier selection metric,proof-of-utility,to enable a lightweight permissioned blockchain.Security analysis shows that our proposed scheme can achieve security and privacy protection.Numerical results indicate that the proposed caching scheme significantly outperforms two benchmark policies in terms of system utility.