Data mining techniques applied to increasing aircrew situational awareness

Abstract

Situation Awareness (SA) is a fundamental concept in the field of human factors in aviation. In the case of aircraft pilots, the most relevant models of SA focus on studying the individual perception of the critical aspects that influence decision making in complex and dynamic environments, i.e., during the flight. In this thesis, we have attempted to provide a comprehensive summary of the factors that surround information management in the cockpit of the aircraft, with special attention to the management of information in air navigation. The research has tried to analyse existing SA models, which already take information management into account, and adapt their interpretation so that the parameters can be measured using Bayesian networks, ultimately intending to provide an estimation of SA. Throughout the research, we have carried out experiments on different machine learning tasks, such as discretization, regression, clustering, and therefore we have improved our knowledge about how to process aeronautical data of various types. We have managed data from several sources with multiple computer tools, using various types of databases, different information exchange formats, testing cloud environments, and successfully sharing information between different applications, including flight simulators. We have developed a set of tools for collecting data in simulated flights, to analyse them, and to estimate models (dynamic Bayesian networks) able of computing and online probability of the SA conditioned to the observations. The thesis also contains the results of the situation awareness estimation.