Modeling and Optimization for Non-Terrestrial Networks

Non-terrestrial networks constitute a vertical scaling of conventional radio ecosystems wherein the optimized hierarchical interplay among unmanned aerial vehicles, high altitude platforms, and satellites is envisioned to augment the coverage and service capabilities of terrestrial networks. The 3D mobility and maneuverability of Unmanned Aerial Vehicles (UAVs) render them principally important non-terrestrial assets. Ergo, this work primarily investigates the functional paradigms of UAV augmented networks across numerous application domains while being supplemented with contemporary radio access technologies, viz., MIMO and Millimeter Wave (mmWave). Specifically, coupled with accurate mmWave air-to-ground channel modeling to statistically characterize signal propagation behavior in a plethora of radio environments, this work leverages tools from dynamic programming and artificial intelligence to develop distributed optimization frameworks intended to enhance the operational potential of UAV assisted networks. This work first elucidates a mmWave propagation modeling campaign on the NSF POWDER experimental testbed, wherein a measurement prototype involving a unique fully autonomous antenna alignment & tracking platform combined with a custom broadband sliding correlator channel sounder facilitates continuous beam-steered measurements in vehicular communication scenarios (emulating air-to-ground links in UAV aided settings), thereby enabling exhaustive signal propagation investigations. Upon establishing the propagation characteristics of mmWave signals in UAV augmented applications and further enhancing UAV link performance via rate adaptation, the subsequent discussions in this work highlight “use-inspired” research efforts spanning a diverse set of potential applications. In particular, the intelligent spectrum sharing problem in UAV aided military deployments is formulated as a Partially Observable Markov Decision Process and solved via randomized point based value iteration; the distributed orchestration of UAV relays in precision agriculture is facilitated by a Semi Markov Decision Process construction, solved using value iteration and subgradient methods; lastly, the decentralized orchestration of MIMO UAVs for data harvesting in industrial automation is enabled via a multiple traveling salesman problem setup, solved via mixed integer combinatorial optimization, i.e., a graphical branch-and-bound algorithm. Across these “use-inspired” efforts, the proposed solution addresses specific problems relevant to each application, alleviates the drawbacks observed in the current literature, and demonstrates superior performance over state-of-the-art algorithms vis-à-vis computational tractability, user quality-of-service, and UAV energy efficiency.