Research On Optimization Strategies Of D2D Assisted Task Offloading And Resource Allocation In MEC System

With the increase in the number of network terminals in mobile communication systems,data volumes have exploded.Traditional Mobile Cloud Computing(MCC)transfers large amounts of data to the cloud data centre resulting in a huge burden on the core network,while the limited backhaul link bandwidth causes network congestion,making it difficult for the cloud data centre to meet the real-time data transmission and computing needs.In addition,smart terminals are limited by resources such as storage capacity,processing power and insufficient energy consumption of chips,making it difficult for terminal devices to meet the needs of computation-intensive applications in terms of arithmetic power and resource requirements.However,Multi-Access Edge Computing(MEC)can bring cloud services down to the edge of the network,converging core capabilities such as network,storage,computing and applications on the edge side of the network closer to users,effectively alleviating the problem of excessive data volume and network congestion in cloud data centres,while compensating for the shortage of resources in mobile terminals.In addition,in 5G communication technology,device-to-device(D2D)communication enables direct link communication between users,which can effectively reduce the load on base stations and improve transmission rates and spectrum utilization.To address the above issues,this thesis proposes three task offloading and resource allocation strategies.The main work is as follows.(1)Two resource allocation and computation offloading strategies under a single-server MEC-D2D system are proposed for different types of application tasks with different demands on latency and energy consumption.Firstly,an optimisation algorithm for joint average time and resource allocation is proposed to maximise the use of MEC server computing resources while minimising the average task completion time.Secondly,an optimisation algorithm for joint average task completion time and energy consumption is proposed to minimise task offloading energy consumption by seeking the computational resource allocation factor of the MEC server.Finally,the simulation results show that the proposed scheme significantly reduces the average task completion time and energy consumption compared to the benchmark scheme.(2)A matching algorithm for multi-server MEC-D2D task offloading and resource allocation is proposed to address the problem of high task cost due to low utilization of hardware resources of idle devices.By deriving a closed-form expression for the computational resources of the MEC server and optimising the user’s transmit power at the same time,the optimal task offloading decision scheme is obtained according to the optimisation objective and the proposed Gale-Shapley-based matching algorithm to effectively reduce the total cost of task execution.Compared with other benchmark schemes,the proposed Gale-Shapley based matching algorithm can effectively reduce the total cost of task execution.(3)A collaborative task offloading and resource allocation algorithm based on task partitioning ratio at the cloud-edge end is proposed to address the problem of high task cost due to high computational task volume.Based on the multi-server MEC-D2D task offloading and resource allocation algorithm,the optimal task partition ratio from the edge end to the cloud end is obtained through theoretical derivation and then transmitted to the cloud data centre for processing.At the same time,a matching algorithm is used to obtain the optimal task offload decision and the optimal transmit power for the user based on the optimal split ratio,thus reducing the system execution task cost.Compared to the multi-server MEC-D2D task offloading algorithm and benchmarking scheme,the proposed cloud-edge-end collaborative task offloading and resource allocation algorithm can further reduce the task execution cost.