A reverse engineering methodology for the development of belief-desire-intention agents

Model-driven architectures, used for the development of Belief-Desire-Intention (BDI) agents, are mostly constructed by following a top-down approach in which software agents are modelled according to a metamodel and the prepared models are given into a series of model-to-model and model-to-text trasnformations to automatically generate the agent implementations. However, modifications made in these auto-generated artifacts can not be reflected back to the agent models. Hence the synchronization between the agent models and the corresponding software is ruined which also makes the reusability of agent models and its software difficult. In order to eliminate these deficiencies, a reverse engineering methodology and a supporting tool for the automatic generation of BDI models from existing agent software are introduced in this thesis. With the use of the proposed tool, it is possible to both re-generate BDI agent models from the programs running on the well-known JACK platform and reflect the changes made in the programs to the corresponding models.