Songwei Li

Graduation Semester and Year

2019

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Electrical Engineering

Department

Electrical Engineering

First Advisor

Yan Wan

Abstract

Unmanned aerial vehicles (UAV) have found broad civilian applications. However, existing commercial usages are limited to single UAVs. To facilitate commercial multi-UAV applications, robust UAV-to-UAV communication with long-distance and broad-band capabilities is critical. Such a communication architecture should not rely on ground infrastructure support, and hence can be applied whenever needed in scenarios where infrastructures are not available. It is foreseen to have extensive usages in a wide spectrum of commercial UAV applications such as emergency response, remote sensing, and nondestructive health monitoring. A promising solution for such a communication architecture is aerial communication using directional antennas (ACDA). We developed several preliminary ACDA solutions. This dissertation presents improvements to the ACDA system, in terms of theory, implementation, and applications. With respect to theory, a critical component of the ACDA system is the automatic alignment of directional antennas to maximize communication performance. To account for an unstable global position system GPS environment, we develop a novel stochastic optimal control solution that integrates reinforcement learning (RL), an effective uncertainty evaluation method called multivariate probabilistic collocation method (MPCM), and an unscented Kalman filter (UKF) for the nonlinear random switching dynamics. Part of this theoretical solution is implemented. With respect to implementation, we redesign the hardware and software to improve robustness, throughput and endurance. Two versions of the ACDA system are implemented. The first version includes a complete new design of platform, communication, computing, control, middleware, and interface components. It features a communication and control co-design to achieve robust performance in a GPS-denied environment. The second version is upgraded with a UAV platform, computing component, battery solution, and rotation structure to reduce interference and improve endurance. Beyond the visual line of sight UAV control using the ACDA system delay is also implemented. With respect to applications, built upon the ACDA system, we first develop a UAV-carried vision-based monitoring system that allows a UAV to continuously track and monitor a mobile infrastructure and transmit back the monitoring information in real-time from a remote location. We then develop a leader-follower tracking system using the ACDA system, that enables cooperative UAV control over a long distance, which can be used in broad multi-UAV remote sensing, monitoring, and emergency response applications.

Keywords

Aerial network, Directional antenna, Reinforcement learning, ROS, Unmanned aerial vehicle

Disciplines

Electrical and Computer Engineering | Engineering

Comments

Degree granted by The University of Texas at Arlington

Download

Share

COinS