Resilient ultra dense networks under UAV coverage for disaster management

Abstract

Resiliency in communication networks is the maintainability of the communication functionality at acceptable levels against possible errors, environmental problems, network outage due to technological causes or malicious attacks. However, it is tremendously time-consuming to redesign the network in a versatile disaster situation considering today's static and conservative communication network infrastructures. In disaster management; assessing the situation, taking immediate and effective precautions and proposing solutions for the optimization is only possible with a robust communication network infrastructure. Additionally, in the case of base station failures, there is no infrastructure to manage the mobile traffic in today's mobile network provider systems. In order to solve this problem, mobile data traffic should be managed adaptively. In recent years, the disasters which were caused by climatic changes cannot be prevented. In case of a natural disaster, the most important thing is to save people's lives. In a situation like this, the first 72 hours is crucially important to react immediately and this can only be possible with quick and effective search and rescue activities. On the other hand, the lack of awareness and communications will vitiate these activities. For example, after the 2011 Tohoku Earthquake in Japan, most of the base stations became out-of-service and the Internet became available barely after 7 days. Also, the service quality of the satellite phones, which are used only for voice communications by the save and rescue teams had decreased. A similar situation was experienced in 2010 after the earthquake in Haiti and long-term communication problems arose due to damaged service provider infrastructures. Similar natural events in the different habitats of the American continent are repeated every year. In addition, after the Marmara Earthquake in 1999, the communication networks have become completely damaged and unusable. In such cases, the continuity of communication is important. In the mentioned disaster scenarios, it is not possible to meet the data demands with the limited physical resources of the infrastructure along with the damages in the existing wireless communication infrastructure. To this end, novel applications are needed in order to solve the network management problems in case of an unanticipated failure. Today, with the increasing use of Unmanned Aerial Vehicles (UAV), many new applications are emerging in the communication sector. According to the Association for Unmanned Vehicle Systems International (AUVSI) Report, direct economic impact from the UAV industry in US is about 3.6 Billion Dollar in 2018 and is expected to exceed 5 Billion Dollar by 2025. In this thesis, UAVs are proposed to support the communication infrastructure as Aerial Base Stations (ABS) via a centralized controller to solve the problems for existing network infrastructures. ABSs have become a promising tool for post-disaster communications. ABS deployment assists terrestrial networks to minimize the disruptions caused by unexpected and temporary situations. Thus, it is aimed to design a resilient network management mechanism with ABSs. ABSs which will be located instead of the failed base stations are advantageous because they have low production and maintenance costs, they have error/damage tolerance and they can easily be controlled and located where humans have limited reach. However, because of the physical limitations with low-capacity power supplies, they have limited flying time, limited velocity and communication range. Moreover, the majority of energy consumption in aerial networks is not spent on computing or communication, but on the power required by engines and flying aerial vehicles. For all these reasons, there are various problems in the system design while trying to accomplish real-world problems and complicated duties. Therefore, in order to increase resiliency in aerial networks, a proper positioning management and a flight planning mechanism are both needed considering the relationship between ABS flight characteristics and energy consumption. Considering the stated reasons, we first focus on on-demand communication. Since on-demand communication can change over time and be hard to accurately predict, it needs to be handled in an online manner, accounting also for battery consumption constraints. This thesis presents an efficient software-based solution to operate ABSs by meeting these requirements which maximizes the number of covered users, and a scheduler which navigates and recharges ABSs in an energy-aware manner. To this end, we propose an energy-aware deployment algorithm and use an energy model to analyze the power consumption and thereby, improve the flight endurance. In addition, we evaluate a novel scheduling mechanism that efficiently manages the ABSs' operations. Our simulations indicate that our approach can significantly improve the flight endurance and user coverage. In the second part of the thesis, we consider that the continuity of the service has increased the challenge of providing satisfactory quality of service. The limited battery capacity and vertical movement with direction switching of ABSs result in frequent interruptions with additional problems related to increased interference, handover delay, and failure of the handover procedure. Therefore, the main goal is to model dynamic mobile network topology and create a scalable structure to manage possible handover procedure between ABSs. With this idea, a solution is presented in a flexible and centralized structure, which analyses the resiliency of the network and is sensitive to increased mobile data traffic and dynamic topology changes. We address the handover procedure in aerial networks by integrating a reinforcement based Q-learning framework. The proposed model enables to ABSs to learn the optimal deployment exploring a Temporal-Difference (TD) learning prediction method. Our study gives a centralized handover procedure avoiding additional overhead to the ABSs and the transition probabilities are estimated to decrease the risk of the handover failure ratio.