BDIE architecture for rational agents
Huiliang Zhang, Huang Shell Ying · 2006
Several successful rational agents have been developed based on the belief-desire-intention (BDI) model In these agents, output actions are usually limited to be primitive. The whole calculation processes of getting the actions must be represented in the plans. Thus, for complex cases, plan library may be hard to create and maintain. This will prevent further applications of the model. In this paper, we will introduce a belief-desire-intention-experience (BDIE) model, which incorporates experience functions into the basic BDI model By this model, some successful algorithms can be involved into the plans of the agents. This provides a way to enrich the agent's abilities without increasing designing difficulty. As an example, a vessel agent is created using this model reinforcement learning and global path planning algorithms are involved in the agent's decision of navigation. The experiment results show that the agent is able to work well with the experience functions.