Applications of uncertainty quantification in the spatial analysis of trajectories

• Main Author: Eerland, Willem Johannis
• Other Authors: Sobester, Andras
• Published: University of Southampton 2017
• Online Access: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.729742

This dissertation looks at analysing security threats via passively guided rocket based weapons of important infrastructure, e.g., airports, military bases, and power stations. It also examines the vulnerability of flight routes, specifically near the airport where the aircraft fly at a relatively low altitude. A key component of the applications presented in this dissertation is quantifying uncertainty, such that scenarios where deviations from a plan occur can be analysed and associated risks can be mitigated. First, a method is proposed to capture motion patterns found in trajectory data. To achieve this, all data are assumed to be generated from a probabilistic model that takes the shape of a Gaussian process. By relying on the Gaussian process framework, the method is able to handle noisy and missing trajectory data. Aircraft trajectory data measured in the vicinity of various airports are analysed via the proposed method. In these examples, flight corridors are visualised and the probability of conflict based on the structure of the corridors is quantified. Furthermore, a strategy aimed at performing the large scale analysis of security risks via a terrain map is introduced. This strategy has shown to improve the detection rate of security threats by at least 20% and detect all risky locations in half the time compared to a brute-force approach. Finally, an open-source stochastic, six-degrees-of-freedom rocket flight simulator is introduced. This simulator assists with the conceptual design of sounding rockets, and produces confidence bounds for a landing location. The uncertainty quantification of the landing location is expanded to the entire flight by capturing the produced (by the simulator) trajectory data in a probabilistic model via the proposed method. These applications lead towards a data driven approach to the analysis of security risks of important infrastructure.