| The emergence of Beyond 5th Generation mobile systems(B5G)and 6th Generation mobile systems(6G)has led to a period of rapid development in mobile communication technology.Mobile Edge Computing(MEC)is being widely recognized as a potential solution and Unmanned Aerial Vehicles(UAV)to be suitable MEC server carriers due to their excellent flexibility is considered.Given that tasks are not independent but interconnected.The collaboration of multiple UAV to provide computational services is believed to align with practical scenarios and hold significant research value.Consequently,the reduction of latency when multiple UAV provide computation services to users has become a popular research direction.At the same time,there is a growing emphasis on green communication.Therefore,while considering the reduction of user task latency.It is also necessary to address the energy consumption throughout the task computation.In summary,this thesis primarily focuses on the following research aspects.The reduction of latency and energy consumption when computation services are provided to users through the collaborative efforts of multiple UAV.Firstly,a three-layer MEC network architecture is proposed,which includes a space-based network formed by multiple UAV,tethered balloons and user terminals.Real-time acquisition of UAV status is enabled by tethered balloons.When a user has task requirements.The tasks are divided into multiple sub-tasks at the user terminal.Task requests for these sub-tasks are then sent to the tethered balloons,which based on the computational resources and locations.The other status information of the UAVs inform the user terminal about which UAV should be utilized for computation of each sub-task.Secondly,the Deep Deterministic Policy Gradient(DDPG)algorithm is employed to optimize the entire computation delay of user tasks.This optimization includes the offloading of tasks,computation and transmission back to the user terminal and result processing of sub-tasks.The analysis of latency caused by different proportions of bandwidth and computational resources considers the allocation of bandwidth during task offloading and the allocation of computational resources’ on MEC servers during computation.Experimental results demonstrate that the adjustment of allocation proportions of bandwidth and computational resources can lead to a reduction in computational latency for user tasks.Finally,the interrelated nature of tasks is considered.The training process is modeled using a Sequence-to-Sequence(S2S)model and the optimization of energy consumption during user task computation is addressed.In this thesis energy consumption is optimized by varying the number of segmented sub-tasks and the task transmission rate.Additionally,the definition of user service quality is employed as a weighted sum of computation latency and energy consumption.By adjusting the number of sub-tasks and the transmission rate.The optimization aims to improve user service quality.Experimental results indicate that different quantities of sub-tasks and transmission rates result in varying levels of energy consumption,weighted sums of latency and energy consumption.Consequently,the adjustment of the number of sub-tasks and the transmission rate can lead to a reduction in energy consumption for user tasks.It improvement in user service quality by considering both latency and energy consumption.In summary,the optimization of latency and energy consumption when computation services are provided to users through the collaboration of multiple UAV is explored in this thesis.Multiple UAV collaborative approach is adopted.The interrelatedness of user task requests is considered.And the effective utilization of bandwidth and computational resources to optimize task latency.Furthermore,the impact of sub-task quantity and task transmission rate on energy consumption is considered.Thereby optimizing user task energy consumption and improving user service quality by comprehensively considering latency and energy consumption. |
