A Controllable Reputation BDI Model

The architecture and reasoning mechanisms of traditional BDI (Beliefs-Desires-Intentions) agent have some vital shortcomings such as the inability to provide a clear and simple decision-making mechanism, and the difficulty to predict and control Agent’s behavior as a result of its autonomy and the lack of conflict management mechanisms, and lack of reliability evaluation of Agent behaviors. These shortcomings fundamentally restrict and hamper BDI based agents applying to some new areas. In this paper, by taking advantage of the non-monotonic knowledge representation and reasoning mechanisms of defeasible logic, a reputation-oriented Agent model is proposed, which is capable of accepting policy guidance, the real-time rule modifications, and handling the run-time rule conflicts. This agent is both autonomous and controllable, and