DATA-DRIVEN AND CONTROL THEORETIC APPROACHES FOR REAL-TIME SENSOR DATA TRANSMISSION OVER UNMANNED AERIAL SYSTEMS NETWORKS
Unmanned Aerial Systems (UAS) are often used to collect and transmit sensor data (e.g., video, radar images) to the ground. While much research related to data transmission in UAS settings has focused on short distances, there is growing interest in operating UAS beyond Visual Line of Sight (VLOS), a relatively unexplored research area. In this thesis, we make three contributions. We present one of the first characterization studies of UAS network performance when operating at distances exceeding VLOS. Our results confirm challenges owing to wireless network variability but also point to opportunities to exploit the correlation of network performance with flight path (distance and orientation). Second, motivated by our
observations, we design Proteus, the first system for video streaming targeted at long-range UAS settings. Proteus is distinguished from existing algorithms developed for traditional Internet settings by explicitly accounting for dropouts, and leveraging flight path information. Through flight emulation experiments, we show Proteus reduces rebuffering from 14.33% to 1.57% at long-range distances, while significantly improving composite video delivery metrics. Third, we design Chimera, which uses the flight path to optimize heterogenous sensor
data transmission. Chimera is based on an optimal control framework, performing online optimization to yield a feedback control policy that makes transmission decisions. Through
emulation and simulation experiments, Chimera reduces penalties related to dropped radar images by 72.4%-100%, compared to an algorithm agnostic of flight path, and achieves an average bitrate of 90.5%, compared to an optimal scheme knowing future throughput, with only minimal increase in radar images dropped.