Multi-UAV Edge Computing Service Assurance For Smart City

With the gradual commercialization of 5G and the wide application of various artificial intelligence(AI)technologies,the Internet of Things(Io T)network continues to expand and has been integrated into all aspects of our lives.The increase of Io T terminal devices makes people’s demand for communication and computing resources continue to increase.In the face of the explosive demand for various resources in the current society,mobile edge computing(MEC)technology has attracted attention and great development by virtue of its characteristics of providing services nearby.Although the deployment of edge base stations and remote clouds solves the task processing and communication requirements of terminal devices to a certain extent.However,people always have unpredictable demands for communication and computing resources at different times and places.For example,the holding of large-scale events or the occurrence of car accidents may lead to the aggregation of terminal devices and a sudden increase in the demand for communication and computing resources.Considering the advantages of flexible deployment,low cost and easy expansion of unmanned aerial vehicle(UAV),UAV edge computing Io T network(UECIN)using UAV to provide fast communication and computing services has become a promising solution to the above problems.In the existing works,there have been many studies on the UAVs’ location deployment,trajectory optimization,resource allocation,etc.,and they are also oriented to various application scenarios,such as post-disaster rescue,smart farms and forests fire protection.However,most of the existing works consider the free space communication between UAVs and ground equipment,and do not consider the influence of complex environmental factors on the ground.Thus,how to deploy UAVs in the face of complex street environments is a challenging problem.In addition,most of the existing works propose short-term service provisioning systems with a fixed number of UAVs,ignoring the problem of UAVs’ limited battery power and the possible changes of ground users’ number,locations and resource requirements.At present,the development of machine learning provides more reliable pre-operation for the autonomous control of UAV systems,which has a favorable impact on the rapid adaptation of UAV groups to changes in ground demands.This paper studies the above-mentioned UAV deployment problem and UAV long-term service problem.The main research contents and innovations are as follows:1.Aiming at the random distribution of ground users,a fuzzy C-means(FCM)clustering algorithm based on large-scale path loss components is proposed to determine the deployment position of UAVs.By comparing with the traditional clustering algorithms based on Euclidean distance,it is confirmed that our algorithm is usable in terms of clustering effect and convergence speed,especially for the highly variable equipment such as UAVs.Then,a UAV deployment scheme is proposed for the Manhattan street scene.Considering the tall buildings around the road,we discuss line-of-sight(Lo S)and non-line-of-sight(Nlo S)communications between the UAVs and ground vehicles,and further optimize the UAV deployment position.Compared with the traditional unsupervised learning K-means clustering algorithm and the FCM algorithm,the simulation results prove that our algorithm has a favorable impact on the communication of vehicles on the established roads in the city,and can reduce the vehicular communication time to a certain extent.2.For the long-term service problem of UAV groups,we propose a dynamic UECIN framework with autonomous prediction characteristics,which aims to provide long-term and stable MEC services for ground users in specific areas.Through the application of Software Defined Networking(SDN),the framework can collect user information in the entire ground area.On this basis,the framework can not only dynamically enter and exit UAVs according to the real-time needs of ground users,but also update the location deployment of UAVs according to the distribution of ground users.Further,we also give an efficient load balancing task allocation scheme to improve the task processing process of UAV groups.Compared with the enumeration method and the traditional greedy algorithm,the simulation results verify the superiority of our framework in UAV deployment,number control and task load balancing.