الفهرس | Only 14 pages are availabe for public view |
Abstract Software agents represent an emerging research direction that aims at automating user tusks. A central task of an agent is planning to achieve its goals. Unexpected disturbances occurring in the agent execution environment represent a serious challenge for agent planning. In this work, a recovery model for the planning process of agents is proposed to cope with these disturbances. The proposed recovery model supports the Hierarchical Task Networks (HTN) planners which represent a broad family of planners that are widely used in agent systems. The correctness of the proposed recovery model is studied and a proof is given for an implementation of HTN planning, namely the Simple Hierarchical Ordered Planner 2 (SHOP2). A prototype for the proposed recovery services is implemented to demonstrate the feasibility of the proposed approach. Of course, robustness in terms or recovery comes with a cost. Therefore, a simulator is built and many simulation experiments were conducted to gain insight about the performance of the proposed recovery model. |