MOTIVATION AND OBJECTIVES. Due to the rapid evolution in wireless technologies and their standardization, wireless communication networks continue to achieve better coverage, higher data rates, support more applications, serve more connected devices, and guarantee strict QoS. Cellular networks have witnessed an unprecedented increase in demand that is perpetually increasing due to the introduction of new services requiring high- speed and reliable connectivity. In the next five years, the number of connected devices is predicted to boom sharply reaching 50 billion [9]; thanks to the new applications and concepts the technology brings. Internet of Things (IoT) and device-to-device (D2D) communications such as connected UAVs and vehicles, and VR are responsible for the exponential increase of new connected devices. As a result, data applications account for the magnitude of traffic generated in cellular networks. Heterogeneous Networks (HetNets) aim at meeting the high data rate requirements of next-generation wireless networks, fifth-generation (5G) communications and provide a solution to the emerging problem. However, this is faced by some challenges that violate the QoS requirements of different services. The number of 5G subscriptions is predicted to reach 190 million by the end of 2020, the majority of which are indoor users [9]. To meet the ever-increasing traffic demand and overcome the challenges HetNets bring, advancement in resource management such as power and rate is crucial.
MOTIVATION AND OBJECTIVES. 3 To date, existing wireless technologies provide limited and perhaps non-practical solutions to overcome the issue of over-crowded networks by ground-user equipment (UE) and aerial users, i.e., unmanned-aerial-vehicles (UAVs). Our work incorporates ML algorithms, namely; Reinforcement Learning (RL) and Distributed Reinforcement Learning (DRL) algorithms into wireless networks to manage resources such as power, rate and spectrum, to find a ubiquitous yet effective solution. To exploit next- generation cellular networks, HetNets provide the needed capabilities to meet the increasing demand by providing adequate solutions via efficient resource allocation. New services and applications have different requirements, such as ultra-reliable and low-latency communications (URLLC) that require high reliability, security, and demand very low network latency. Additionally, new technologies allow diverse industries to develop new models and deploy various logical networks over one infrastructure. Two complementary but independent concepts; Software-defined Networking (SDN) and Network Function virtualization (NFV) are introduced to meet the stringent requirements of 5G different use cases, Enhanced mobile broadband (eMBB), URLLC and Massive machine-type communications (mMTC). With the advancement in network developments, the increase in traffic loads and business dependency to their networks, SDN and NFV allow for flexibility and automation. Besides, virtualization is essential to upgrade new network functions and locate them where and when needed in the network. Depending on area and coverage radius, a smallcell (e.g., femtocell) can be overlaid on the existing infrastructure. In our research, we only consider femtocell as we focus on indoor UEs that are usually served by a user-deployed Femto Access Point (FAP). FAPs are practical due to their cost efficiency, ease of deployment and cover short ranges, making them not scars to losses and interference. Two-tier or multi-tier HetNets principally comprise of a macrocell and overlaid smallcells such as femtocells, picocells, and relays, inevitably aiming at enhancing the coverage and capacity of existing cellular networks; to meet the growing demand of data traffic. Smallcells play a crucial role in supporting congested macrocells by offloading some of the traffic by increasing the overall cell coverage [10]. The primary objective of this research is to develop and investigate novel resource allocations frameworks and ...
MOTIVATION AND OBJECTIVES. Despite the many impressive examples of human inventiveness, our technological advances for management and control of Technological System-of-Systems (TSoS) are still pale in comparison with the elegance, effectiveness and supreme functionality found in Natural System-of-Systems (NSoS) such as e.g., biological systems, the human brain, animal herds (swarms), teams of interacting/cooperating humans and ecological systems. The human developments, systems and methodologies for controlling the behaviour of TSoS, such as, e.g., transport and energy/water systems, emergency crisis management systems and complex industrial production systems, are by no means comparable to the "wisdom and effectiveness" of NSoS. In NSoS, a large number of constituent systems - operating, interacting, adapting, evolving and self-organizing in a fully- autonomous and local manner - achieves a highly efficient, self-adjusting, self-organizing and dependable performance of the overall NSoS. Even in cases where the NSoS constituent systems possess a highly heterogeneous nature and are mutually interacting and communicating through a very complicated and constantly changing topology, environment and hierarchy, the emerging NSoS is enabled to very effectively profit from each constituent system's "speciality" and comparative advantages towards successfully and rapidly meeting the overall NSoS1 goals and objectives. What mostly characterizes NSoS is that local behaviour and interaction of the constituent systems leads to satisfaction of the NSoS goals at the global level without the need for a supervisor that gives direct orders and tells each constituent system what to do. In fact, each constituent system is "an autonomous unit that reacts depending only on its local2 environment" and its own capabilities. In cases where there is a supervisor or leader in the NSoS, it provides only with high-level goals for the overall NSoS, rather than specific commands to the constituent systems. Moreover, as the constituent systems operate in a local and autonomous fashion, the overall NSoS achieves to operate without the need for an "expensive and complete infrastructure", i.e., without the need to provide to each and every constituent system with complete information stemming from all over the NSoS. Most importantly, there is no need for tedious re-design and re-configuration operations whenever one or more constituent systems are to enter or leave the NSoS or when some of the constituent systems ar...
MOTIVATION AND OBJECTIVES. Despite the impressive methodological and technological advances in the field of underwater robotics in recent years, the existing approaches are far from being able to design and deploy teams of Autonomous Underwater Vehicles (AUVs) that can fully autonomously take over real-life complex situation awareness operations such as environmental monitoring and clean-up operations, seafloor mapping, security and surveillance, inspection of underwater structures, etc. In the majority of such applications, ROVs (Remotely-operated Underwater Vehicles) are employed, where human operators are called to (a) “assess and understand” the ROV’s surrounding environment via the information received by the vehicles’ sensors and (b) remotely navigate the vehicles in an attempt to accomplish the particular mission the vehicles are deployed for. The involvement of human operators is very costly, as it requires well-trained and highly-experienced personnel for long periods of time, and, in most cases, it renders the deployment of cooperative systems of underwater vehicles infeasible as it is practically impossible for the operators to concurrently monitor/navigate the ROVs and coordinate with other operators. Very recently there has been an intensive research and development effort towards designing and deploying teams of cooperative AUVs that overcome the above-mentioned shortcomings of humanly-operated ROVs: various R&D projects (including the FP6 ICT projects GREX and VENUS and the FP7 ICT projects Co3-AUVs, TRIDENT, C4C and SHOAL) have developed or are developing new methodologies and designs for multi-AUV1 coordination and cooperative control. These projects seek to develop efficient algorithms for AUVs operating under severe environmental conditions (e.g. in the presence of strong currents and turbulences). Moreover, they need to deal with the poor communication, sensing, and localization capabilities of underwater vehicles which mainly rely on sonars both for communications and sensing, while navigating in GPS-denied or GPS-limited environments. Very impressive test cases have been (or will be) demonstrated as parts of the aforementioned R&D projects, with teams of AUVs being able to perform – cooperatively – formation control, path planning and target tracking, seafloor exploration and environmental monitoring, underwater object manipulation, etc. However, despite the advances made through current multi-AUV R&D endeavors, the existing or planned multi-AUV systems are far...
