Evaluation of Similarity Measures for Flight Simulator Training Scenarios

Abstract

Flight simulator training is fundamental for the acquisition and maintenance of professional pilot skills. One of the key factors for the effectiveness of this type of training is the design of the scripts of the sessions, usually called “scenarios”. Currently, civil aviation authorities are advocating a customization of the flight training scenarios based on the specific needs of each pilot, which makes their creation a very demanding task in time and resources. Automatic generation systems for these scenarios have been proposed in the scientific literature, but they have not been fully applied to commercial flight simulators yet. In this paper, we review the most important advances in this field to date and introduce a first proposal of a case-based reasoning system for the generation of training scenarios for non-technical skills. Particularly, our goal is to evaluate a set of four different similarity measures for case retrieval of event sets found in these training scenarios, using the judgement of real experts in the field as validation method.