Abstract

Since aircraft capabilities and information complexity have increased, flying an airplane has become more complex. Consequently, pilots run the risk of information overload and making mistakes. That is why situation awareness is usually considered to be one of the most important aspects when flying an airplane. Thus, a number of systems have been developed to help prevent mistakes while flying by increasing the situation awareness of the pilot. These systems filter the huge amount of information for the pilot and provide him with only relevant information based on the current situation the pilot is in and take into account the workload of the pilot. This report tries to point out how human behavior can be modeled in order to realize an automated pilot (called bot) that acts like a human one. Many automated pilots already exist. However the purpose here is the human resemblance. The system designed for the project focuses on the decision-making process of a bot during a dogfight. First the domain knowledge is described that has been gathered about how a real pilot behaves while dog fighting. That is to say which data it takes into account and how it does it in order to choose and execute the right maneuver. A cognitive model is then presented that defines how information is processed and how decision is taken. A first prototype, which implements this cognitive model, has been realized. It stands as the first step of an iterative approach to improve the behavior of the bot. Decision tree processes are run to take decision. Since the results of the prototype were not good, a second release has been carried out. Many improvements have been added and the decision trees processes have been kept. Running the decision tree processes leads to several possible maneuvers. In this release a way to dismiss incompatible maneuvers has been added in order to elect the most suitable one. The results for this second release are not yet available.