A Runtime Goal Conflict Resolution Model for Agent Systems

The goal-oriented agent programming based on BDI model is obtaining increasing attentions, because it allows us to design proactive behaviors for an agent. Generally, an agent does pursue multiple goals not only in a sequential way, but in a simultaneous mode. Accordingly, a rational agent requires that the goals pursued by it are consistent and do not conflict or even deadlock with each other. Existing strategies of goal deliberation are still not sufficient to deal with this task. This paper proposes a runtime goal conflict resolution model for agent systems, which consists of a goal state transition structure and a goal deliberation mechanism based on an extended event calculus. Besides the elaboration on the theoretical parts, a case study is also given to show how this model works. Finally the evaluation of the model is introduced, and the comparing result shows that the goal deliberation time of the agent can be decreased and fewer goal conflicts could happen in the runtime.