Research On The Resource Management In Internet Of Things Based On Mobile Edge Computing

With the rapid development of the Internet of Things(IoT)applications and the exponential growth of the service data,massive IoT applications are required for realtime data transmission and processing,such as self-driving,smart grid,e-healthcare,etc.However,the cloud computing platform not only increases the end-to-end transmission latency,but also significantly increases the energy consumption.Recently,Mobile Edge Computing and Caching(MECC)technology can perform a large amount of data transmission,information integration and optimization by migrating the cloud computing capability to the distributed edge networks.The advantages of MECC can reduce the service latency while reducing the system cost.In the complicated IoT application scenarios,service requirements and resource availability are a continuous process.Massive IoT devices need to frequently to perform a series of operations,such as communication,computation and caching.Considering the IoT devices with limited battery power,these operations would increase the interruption probability of data transmission and computation.On the other hand,the inherent characteristics of the Internet of Vehicles(IoV)pose great challenges to computation offloading and resource management during data transmission.In addition,the edge access network does not utilize its nature advantages close to the data source.It lacks the capability to perceive service requirement,the device feature and resource availability.In order to provide more intelligent decision-making function,it is necessary to analyze the more features of edge network,so as to satisfy the dynamic matching among service requirements,equipment features and resource states.This thesis focus on integrating the MECC and artificial intelligence technologies to address the aforementioned problems in terms of task offloading,energy scheduling,access mode selection,mobile-aware content caching,distributed resource trading and reliable computing.The thesis is mainly divided the following four chapters: 1)Joint task offloading and energy scheduling in device-todevice edge computing networks;2)Collaborative computation offloading in vehicular edge multi-access computation networks;3)Mobility-aware caching in vehicular edge caching networks;and 4)Blockchain empowered distributed computation offloading.In the first part,we design a device-to-device edge computing network(D2D-ECN),a new paradigm for computation offloading and data processing with a group of resourcerich IoT devices.Energy harvesting technology is utilized to ensure the continuity of data transmission and processing for the possibility of computation interruption in devices powered by battery.In order to achieve the good balancing between service latency and energy consumption,a joint optimization of task offloading and resource allocation is modeled as a constrained markov decision process(CMDP).Considering the randomness of renewable energy and channel state,two different learning algorithms are proposed to deal with the challenges that the traditional optimization schemes cannot directly solve the above the CMDP problem.In the point-to-point offloading system,the computation offloading algorithm based on learning can interact with the external environment and obtain the optimal offloading strategies to minimize the objective.On the other hand,this paper proposes an online offloading algorithm based on Lyapunov optimization to tackle with the curse of dimensionality and the problem of continued action space.Without the priori network information,the Lyapunov stability proves how the CMPD problem can be transferred into a joint optimization problem of task offloading,energy scheduling and computation resource allocation under a single period.Furthermore,a meta-heuristic algorithm is used to solve the optimization problem in low computation complexity.Compared to the existing algorithms,simulation results demonstrate the proposed offloading algorithm can significantly reduce the energy consumption and the service latency.In order to solve the problem that the price mechanism oriented resource allocation cannot motivate vehicles to share traffic information in real time,the second part introduces the preference utility of vehicles and edge servers.A computation offloading mechanism based on two-side matching theory is proposed to maximize the utility of both vehicles and edge servers.Different from the traditional method to solve the mixed integer programming,the proposed algorithm is a stable association between vehicles and edge severs according to their own preference list.On the other hand,this paper proposes a joint optimization algorithm for access mode selection and task assignment in heterogeneous IoV,aiming to maximize the sharing gain of transportation while meeting the diverse requirements.Aiming at this complex mixed integer programming problem,we propose a collaborative offloading scheme based on deep reinforcement learning,which can well overcome the challenge of curse of dimensionality by the increase of network states of traditional learning.Compared withe non-cooperative computation offloading scheme,numerical results show that the proposed algorithm can reduce the service latency and energy consumption.The third issue proposes a cooperative edge caching scheme,a new paradigm to jointly optimize content placement and content delivery in vehicular edge computing and networks,with help of flexible trilateral cooperation among macro-cell station,roadside units and smart vehicles.We formulate the joint optimization problem as a double time-scale Markov decision process(DTS-MDP),based on the fact that the time-scale of content timeliness changes less frequently compared to the vehicle mobility and the changes of network states in the content delivery process.At the beginning of the large time-scale,the content placement/updating decision can be obtained according to the content popularity,vehicle driving paths and resource availability.On the small time-scale,the joint vehicle scheduling and bandwidth allocation scheme is designed to minimize the content access cost while satisfying the constraint on content delivery latency.To solve the long-term mixed integer linear programming(LT-MILP)problem,we propose a nature-inspired method based on deep deterministic policy gradient(DDPG)framework to obtain a suboptimal solution with a low computation complexity.Simulation results demonstrate that the proposed cooperative caching system can reduce the system cost,content delivery latency and improve content hit ratio,compared to the non-cooperative and random edge caching schemes.The forth part accounts for key challenges of D2D-ECN in terms of the efficiency of the resource management and the resulting security concerns caused by lacking trustworthy between task owners and resource providers.In particular,we propose to use a blockchain-empowered framework for implementing resource trading and task assigment as the smart contracts.However,the existing Proof-of-Work(PoW)is impractical for the resource-constrained IoT devices due to high computational complexity of the mining process.Thus,we present a reputation-based consensus mechanism called proofof-reputation(PoR),where the device with the highest reputation score is responsible for packaging the resource transactions and reputation records in the blockchain.Furthermore,we evaluate the reputation score of each device according to the current computation performance and history reputation.Security,feasibility analysis and numerical results show that our proposed computation offloading scheme can be deployed in the decentralized D2D-ECN system safely and effectively.