Predictive Analytics Lead to Smarter Self-Organizing Directional Wireless Backbone Networks

Abstract

Directional wireless systems are becoming a cost-effective approach towards providing a high-speed, reliable, broadband connection for the ubiquitous mobile wireless devices in use today. The most common of these systems consists of narrow-beam radio frequency (RF) and free-space-optical (FSO) links, which offer speeds between 100Mbps and 100Gbps while offering bit-error-rates comparable to fixed fiber optic installations. In addition, spatial and spectral efficiencies are accessible with directional wireless systems that cannot be matched with broadcast systems. The added benefits of compact designs permit the installation of directional antennas on-board unmanned autonomous systems (UAS) to provide network availability to regions prone to natural disasters, in maritime situations, and in war-torn countries that lack infrastructure security. In addition, through the use of intelligent network-centric algorithms, a flexible airborne backbone network can be established to dodge the scalability limitations of traditional omnidirectional wireless networks. Assuring end-to-end connectivity and coverage is the main challenge in the design of directional wireless backbone (DWB) networks. Conflating the duality of these objectives with the dynamical nature of the environment in which DWB networks are deployed, in addition to the standardized network metrics such as latency-minimization and throughput maximization, demands a rigorous control process that encompasses all aspects of the system. This includes the mechanical steering of the directional point-to-point link and the monitoring of aggregate network performance (e.g. dropped packets). The inclusion of processes for topology control, mobility management, pointing, acquisition, and tracking of the directional antennas, alongside traditional protocols (e.g. IPv6) provides a rigorous framework for next-generation mobile directional communication networks.

This dissertation provides a novel approach to increase reliability in reconfigurable beam-steered directional wireless backbone networks by predicating optimal network reconfigurations wherein the network is modeled as a giant molecule in which the point-to-point links between two UASs are able to grow and retract analogously to the bonds between atoms in a molecule. This cross-disciplinary methodology explores the application of potential energy surfaces and normal mode analysis as an extension to the topology control optimization. Each of these methodologies provides a new and unique ability for predicting unstable configurations of DWB networks through an understanding of second-order principle dynamics inherent within the aggregate configuration of the system. This insight is not available through monitoring individual link performance. Together, the techniques used to model the DWB network through molecular dynamics are referred to as predictive analytics and provide reliable results that lead to smarter self-organizing reconfigurable beam-steered DWB networks.

Furthermore, a comprehensive control architecture is proposed that complements traditional network science (e.g. Internet protocol) and the unique design aspects of DWB networks. The distinct ability of a beam-steered DWB network to adjust the direction of its antennas (i.e. reconfigure) in response to degraded effects within the atmosphere or due to an increased separation of nodes, is not incorporated in traditional network processes such re-routing mechanism, and therefore, processes for reconfiguration can be abstracted which both optimize the physical interconnections while maintaining interoperability with existing protocols. This control framework is validated using network metrics for latency and throughput and compared to existing architectures which use only standard re-routing mechanisms.

Results are shown that validate both the analogous molecular modeling of a reconfigurable beam-steered directional wireless backbone network and a comprehensive control architecture which coalesces the unique capabilities of reconfiguration and mobility of mobile wireless backbone networks with existing protocols for networks such as IPv6.