Boris Kerkez and Michael T. Cox
The main focus of this research is to establish the techniques and prove feasibility of the case-based keyhole plan recognition dealing with incomplete plan libraries. Most traditional plan recognition systems operate with complete plan libraries that contain all of the possible plans the planner may pursue. However, enumeration of all possible plans may be difficult (or impossible) in some complex planning domains. Furthermore, the completeness of the library may result in occurrence of extraneous plans that may impact the recognizer’s efficiency (Lesh and Etzioni, 1994). The main difficulty when dealing with incomplete plan libraries is the recognizer’s inability to reason about the planner’s intentions that are not contained in the plan library. The recognizer that deals with incomplete plan libraries will only be useful if it is capable of making the intent predictions based on the limited information already present in the library. Most traditional recognition systems reason in terms of planning actions and do not explicitly keep track of the world states visited during the execution of a plan, except for the initial and the goal states. On the other hand, our system reasons in terms of both actions and situations in which the planner finds itself. Planning situations, represented by the states of the planner’s environment, enable the recognizer to form predictions in cases where traditional systems would falter, such as making predictions in light of novel planning actions. Our experimental results demonstrate the effectiveness of the case-based plan recognition with incomplete libraries.