Offloading Optimization Of Mobile Edge Computing Assisted By Unmanned Aerial Vehicles

With the rapid growth of various new mobile terminal devices,the demand for computing resources has surged,and with limited infrastructure coverage,traditional mobile edge computing is facing serious challenges,so mobile terminals need a competitive and scalable access network.With the booming development of drone technology,drones provide a solution to the problem of insufficient computing resources because of their high mobility,low cost and high reliability.However,UAVs have limited flight time due to problems such as limited energy storage.In order to solve the above problems,this paper studies and discusses the problem of UAV-assisted mobile edge computing and UAV flight trajectory optimization.The main work is as follows:(1)For the scenario of insufficient computing resources,a UAVassisted multi-user mobile edge computing system is designed,which mainly consists of an edge server,a remote cloud server,an Unmanned Aerial Vehicle(UAV),and multiple users,where the UAV can provide extensive communication and certain computing power to the users.In order to optimize the task offloading decisions and the allocation of computational and communication resources among all users,an adaptively tuned offloading optimization(USS)algorithm is proposed to achieve the overall energy consumption and latency minimization.Experimental results show that the proposed algorithm can significantly reduce the offloading cost.(2)For the problem of limited UAV energy storage,an adaptive trajectory optimization scheme for multi-UAV-assisted unloading is designed,firstly,a UAV capability evaluation method is proposed to evaluate the capability of different UAVs;then an adaptive clustering method is proposed to cluster the unloading area,which is divided according to the distance and the density of the area calculated by scanning the area,followed by the size of the area task and the capability of the UAV strengths and weaknesses are matched sequentially;finally,the problem is considered as a combinatorial optimization problem,and a pointer network is proposed to plan the trajectory of the UAV,and the parameters of the pointer network are trained with reinforcement learning to minimize the cost of the UAV in the mobile edge computing system and produce a nearoptimal trajectory.Experimental results show that the proposed algorithm can significantly reduce the total system cost.