Research On Deployment And Association Techniques For Unmanned Aerial Vehicle Networks

The uneven distribution of network communication requests in geography and time has grown increasingly pronounced with the continual scaling up of mobile cellular networks,which poses severe challenges for cellular network planning and operation.It is difficult to dynamically adjust the network structure to provide adaptive communication services for users,once the terrestrial base stations in the cellular networks are deployed.As a result,telecom operators have erected an increasing number of terrestrial base stations(such as macro and micro base stations)in order to guarantee the users’communication experience.However,an excessive number of ground base stations not only results in high deployment and maintenance costs but also wastes communication resources in areas where communication requests are only made briefly,such as stadiums and theaters.This will not be conducive to the sustainable development of the communication network.With the widespread application of unmanned aerial vehicles(UAVs)flight platforms,integrating the communication base stations into UAVs and turning them into mobile communication base stations has gotten a lot of attention in recent years.UAVs present significant advantages in terms of mobility and maneuverability and cost.Introducing the UAV base station into the traditional ground communication network and then adjusting UAVs’ deployment locations and quantity according to the spatial-temporal distribution of users can not only support the dynamic communication requests in the cellular network,but also avoid the redundant terrestrial base stations and save the communication resources.In this context,the fundamental problem that has to be solved is how to design the deployment and association strategy of UAV base stations to match the spatial-temporal pattern of ground users and meet their communication requirements in the cellular network.The majority of the previous studies looked at UAV deployment and association schemes based on users’ real-time and exact location data.However,acquiring the user’s real-time and precise position information is quite difficult.Furthermore,the existing networks do not have the capability of actively obtaining real-time and precise user location information.Consequently,this dissertation mainly focuses on the investigation over the multiple UAVs deployment and association schemes based on the statistical user distribution information.The main contributions are summarized as follows.1)A multi-UAV deployment scheme in three-dimensional space is proposed to provide on-demand deployment for ground users,based on user statistical distribution information.Specifically,the three-dimensional UAV deployment problem is divided into two parts:UAV horizontal deployment and UAV vertical deployment.For the problem of horizontal deployment of UAV,a virtual force field is established by uniformly sampling the user distribution.The UAVs that are located in the virtual force field will be affected by attractive and repulsive forces.The final UAV horizontal deployment position can be determined by calculating the force balance point of UAVs in the virtual force field.On this basis,the vertical positions of UAVs are determined by using the classical particle swarm algorithm.In order to verify the effectiveness of the proposed algorithm,the user distribution data in the real-world is introduced.Simulation results show that the proposed algorithm can not only provide the UAV deployment strategy that matches user distribution,but also improve network throughput performance significantly.2)In order to improve the timeliness of the UAV deployment scheme for the scenarios with real-time and dynamic user distribution,a fast and on-demand deployment strategy for UAVs based on the deep convolutional neural network is developed.In this work,the traditional UAV deployment problem has been transformed into a regression problem in the deep learning field.The training set,testing set for the deep neural network are established through using a few user distribution samples and the optimal UAVs’deployment position samples.Then,the loss function of the deep neural network is constructed and the classical Adam gradient descent algorithm is used to optimize the parameters of the deep neural network to minimize the loss function.Simulation results show that the proposed method can predict the UAV deployment position quickly while guaranteeing the accuracy of the UAVs’ deployment positions.3)Considering the cellular network with the ground base stations and UAVs,UAVs forward user communication data to the base station.In order to minimize the required transmit power for communication,a joint design of UAV deployment and association schemes based on reinforcement learning theory and optimal transport theory is proposed.Specifically,the UAV deployment problem is modeled as a Markov decision process,and a multi-agent Q learning-based UAV deployment framework is proposed.In this deployment framework,the UAV action selection mechanism relies on the user association design.In order to obtain the optimal UAV action strategy,we introduce the optimal transport theory to derive the optimal user association scheme,and also prove the uniqueness of the optimal user association scheme.Through joint designing,the on-demand deployment and association scheme of UAV relays is finally developed.Simulation results show that the proposed scheme can not only ensure a balanced network load distribution,but also reduce the transmit power for communications and improve energy efficiency performance.4)The requirement for communication and computation forces UAVs to play the role of mobile edge computing nodes.In this context,UAVs will process the tasks offloaded by users locally or forward user tasks to the cloud server.In order to reduce the energy consumption of communication and computation,a joint optimization strategy for UAV edge computing node deployment,association,task offloading is proposed.Specifically,the original problem is divided into three sub-problems.First,a modified virtual force field-based UAV mobile edge computing node deployment scheme that matches user distribution is studied.After that,the user association strategy and the task offloading strategy of the UAV mobile edge computing node are jointly designed by using the optimal transport theory and the convex optimization theory.Compared with the benchmark scheme,simulation results show that the proposed method can reduce communication and computational energy overhead greatly.In summary,to tackle the shortcomings of the existing schemes,this dissertation investigates a variety of UAV deployment and association methods based on statistical user distribution information,which can greatly improve the communication performance of the UAV network compared with the traditional schemes.