Case-based decision making for diversion scenarios

Problems with decision-making have been pointed out as contributing factors to the occurrence of aeronautical incidents and accidents. In this work, we tackled the decision-making process related to diversion scenarios. Diversion is a deviation or alternative from the original flight plan, usually due to some problem caused in the aircraft, such as leakage of fuel, of engine and in aerodynamic surfaces, among other problems of the aircraft that can make it impossible to arrive at the previously defined destination airport. We propose a case-based reasoning (CBR) model for aiding the pilot to deal with diversion situations, chosen as adequate scenarios from the results of a survey conducted with pilots to identify those for which an AI-based approach might be helpful regarding decision making. A CBR system basically uses a memory of recorded cases for generating new solutions to similar problems, it thus contains a "memory" of cases with descriptions of specific situations that occurred in the past, solutions of these described situations, and the evaluation of their results. The reasoner solves new problems by adapting in-memory solutions to it, via some similarity criteria. Specific similarity measures for comparing cases and retrieving the ones most similar to the diversion situation at hand were defined, and a database for 26 diversion cases was built for conducting the experiments on decision making. Case retrieval experiments over 20 diversion test situations produced a success rate (that is, a diversion suggestion compatible with pilot decisions) of 85%. We intend to perform as future work an extension of the cases database by incorporating real diversion situations,dynamic attributes related to weather and traffic intensity at the diversion airport, and a more comprehensive assessment by pilots, possibly towards a learnable automatic weight adjustment for the attributes that compose the case similarity metric.