Optimization of 3-D location and resource allocation problems of unmanned aerial base stations

Unmanned Aerial Vehicles (UAVs) have started to be used in many areas. Their cost advantages and the ease of manageability allow for a broader range of applications. In this thesis, the integration of UAVs into next generation wireless communication networks as mobile base stations is considered and several UAV Base Station (UAV-BS) location and resource allocation problems are solved under various scenarios. The vertical movement ability of UAV-BSs and non-convex and non-monotone structure of performance measures used in wireless communication networks differ from classical location problems and increase the complexity of problems. Four different problems are considered to optimize UAVBS location and resource allocation decisions. The first problem considers a static uncapacitated generalized UAVBS covering location decisions, the second problem considers a static capacitated single UAVBS location and resource allocation decisions, the third problem considers a static capacitated multi-UAVBS location and resource allocation decisions, and the fourth problem considers a dynamic uncapacitated single UAVBS covering location decisions. These problems are typically modelled as mixed integer non-linear programming formulations, and the properties specific to each model are investigated. All of the considered problems belong to NP-Hard problem class. Therefore, it is not possible to find exact solutions in reasonable solution times by commercial solvers for large problem instances. For this reason heuristic algorithms are developed to obtain quick and efficient solutions. These algorithms mainly depend on determining the location and resource allocation decisions separately and iteratively and improving the final solution by different local search algorithms. Different than the classical location problems, we consider both the network capacity and different user demand behavior in the design of these heuristic algorithms. Since it has not become a standard to use UAV-BSs in the existing communication networks yet, there does not exist any publicly available data set for this problem. Therefore, the performances of the developed formulations and proposed heuristic algorithms are tested on synthetically generated data sets. Each generated data set for each problem is also expected to become a benchmark for future research. Extensive computational tests are performed to compare both the solution quality and time of the formulations that are solved by commercial solvers and the developed heuristic algorithms. In the light of the attained results, several policies are proposed to improve the coverage performance of UAV-assisted wireless communication networks. In conclusion, it is shown that the developed heuristic algorithms find approximate optimal solutions within reasonable CPU time, and that UAV-BSs would substantially increase the performance of wireless communication networks.