Search

A strong emphasis on safety in commercial and military aviation is as old and as significant as the field of aviation itself. With the growing role of autonomy in aviation, the future of flight comprises of two general directions: manned and unmanned. Manned aircraft is the more established area, in which a human flight crew serves as the main driving force in ensuring an aircraft’s safety and success. Within this time-tested concept, the most significant bottleneck of safety lies within a crew managing tasks of high mental workload. In recent years, autonomy has aided in easing cognitive workload. From there, the challenge lies within applying a seamless blend of human and autonomous control based on the needs of one’s mental load. Meanwhile, the field of unmanned aerial vehicles (UAVs) poses its own unique challenges of integrating into a shared airspace and transitioning from remote human-centric control to fully autonomous control. In such a case, minimizing discrepancies between predicted UAV behavior and actual outcomes is an ongoing task to ensure a safe and reliable flight. While manned and unmanned flight safety may seem distinctly different in these regards, this dissertation proposes an overarching common theme that lies within the ability to effectively model inputs and outputs through machine learning to predict potential safety hazards and thereby improve the overall flight experience. This process is conducted by 1) evaluating different machine learning techniques on assessing cognitive workload, 2) predicting trajectories for autonomous UAVs, and 3) developing adaptive systems that dynamically select appropriate algorithms to ensure optimal prediction accuracy at any given time.The first phase of the research involves the manned side of flight safety and does so by examining effects of different machine learning techniques used for assessing cognitive workload. This begins by comparing the different algorithms on four different datasets involving cognitive activity based on physiological and subjective data. From there, two new algorithms are developed that dynamically select a machine learning technique based on the attributes of the given physiological data: one that statically chooses a method and another that dynamically changes methods over time based on which is projected to provide optimal accuracy and efficiency. By being able to accurately classify an activity with a certain amount of expected cognitive load, this can be applied in aircraft to assist pilots in early detection of mental overload and underload.The second phase then invokes unmanned flight safety by aiming to enhance prediction of autonomous UAV data. This is done by fusing navigational coordinates with radar data (e.g. how close a UAV is to another vehicle or an obstacle) using Dempster-Shafer Evidence Theory. The algorithm is then compared against other mechanisms for UAV data prediction with encouraging results. Finally, an additional algorithm is developed that dynamically chooses methods at different points in time based on which is expected to produce the best accuracy. Thus, by improving accurate predictions of future UAV data, unmanned flight safety can be enhanced by minimizing discrepancies between expectations and outcomes in an increasingly unpredictable shared airspace.

Wireless sensor networks are a collection of small, disposable, low-power devices that monitor vital sensory data for a variety of civil, military, and navigational applications. For instance, some cities have a network of emergency phones scattered across walkways so that citizens in distress can immediately reach emergency services. Using effective localization techniques that are both highly accurate and of low computational cost, 911 services can dispatch police, fire, or medical services to a caller's location as quickly as humanly possible. Hence, from the standpoint of locating a node in a network, every percent of accuracy achieved and every second of time saved can be the difference between life and death. This thesis presents two novel algorithms for wireless sensor network localization through the incorporation of Dempster-Shafer Evidence Theory. The first technique follows a verbose methodology for node positioning that fuses multiple types of signal measurements, such as received signal strength and angle of arrival, and utilizes the expected value property of DS Theory to geo-locate a node with a moderate accuracy of 78-87%, thereby providing an introductory approach to the previously untapped fusion of WSN localization and DS Theory.The second approach consists of a low cost, highly accurate data fusion technique that incorporates the plausibility property of DS Theory to establish a high level of accuracy. Due to this unique approach to data fusion and predictive data modelling, this second algorithm achieves an optimal accuracy range of 83-98% in a flexible multitude of simulation scenarios at a fraction of the runtime required under prior established localization techniques. Overall, these two algorithms provide a groundbreaking new application of Dempster-Shafer Theory as well as fast, accurate, and informative new approaches to wireless sensor network localization that can improve a wide range of vital applications.