Abstract

The report describes the design and the implementation of a prototype for a fully autonomous agent, which can fly an airplane during a dogfight. This project is one of the multiple projects that are done within the ICE project of the Knowledge Based Systems group of the Delft University of Technology. Until now, many projects, also within the ICE project, focused on one specific element of flight automation. This project focuses on combining multiple elements of flight automation in one fully autonomous agent. In this project certain methods that are proposed in earlier reports are studied and some of them are used for the dogfight agent. A flexible and modular overall architecture is presented in which, for each element of the flight automation process, different methods can be implemented and replaced easily. The methods that are used for the prototype implementation are presented with a detailed design and a description of the implementation. The prototype is successfully implemented and can really act as a competitive player during a one-to-one dogfight in Microsoft Combat Flight Simulator. An important conclusion is that the performance of the whole agent is dependent on the performance of all the different elements. This conclusion underlines the relevance of this project, because the basis of the project was to split the concept of flight automation into multiple elements, so it is possible to focus on the individual elements.