Backhaul-Aware Optimization of UAV Base Station Location and Bandwidth Allocation for Profit Maximization

TL;DR

This paper presents a joint optimization framework for UAV base station placement and bandwidth allocation to maximize network profit, considering backhaul constraints and using a novel heuristic approach for complex problem solving.

Contribution

It introduces a combined optimization model for UAVBS location and resource allocation, along with an efficient heuristic algorithm to solve the complex MINLP problem.

Findings

The heuristic algorithm outperforms alternative solutions.

The problem reduces to a Multidimensional Binary Knapsack Problem when location is fixed.

Numerical results demonstrate improved network profit with the proposed method.

Abstract

Unmanned Aerial Vehicle Base Stations (UAVBSs) are envisioned to be an integral component of the next generation Wireless Communications Networks (WCNs) by dynamically moving the supply towards the demand. A significant drawback of the state-of-the-art have been designing a WCN in which the service-oriented performance measures (e.g., throughput) are optimized without considering different relevant decisions such as determining the location and allocating the resources, jointly. In this study, we address the UAVBS location and bandwidth allocation problems together to optimize the total network profit. In particular, a Mixed-Integer Non-Linear Programming (MINLP) formulation is developed, in which the location of a single UAVBS and bandwidth allocations to users are jointly determined. The objective is to maximize the total profit without exceeding the backhaul and access capacities. The profit gained from a specific user is assumed to be a piecewise-linear function of the provided data rate level, where higher data rate levels would yield higher profit. Due to high complexity of the MINLP, we propose an efficient heuristic algorithm with lower computational complexity. We show that, when the UAVBS location is determined, the resource allocation problem can be reduced to a Multidimensional Binary Knapsack Problem (MBKP), which can be solved in pseudo-polynomial time. To exploit this structure, the optimal bandwidth allocations are determined by solving several MBKPs in a search algorithm. We test the performance of our algorithm with two heuristics and with the MINLP model solved by a commercial solver. Our numerical results show that the proposed algorithm outperforms the alternative solution approaches and would be a promising tool to improve the total network profit.