Mathias Bauer - Susanne Biundo - Dietmar Dengler
Harald Feibel - Jana Koehler - Gabriele Paul
Wolfgang Wahlster

An Overview

Intelligent help systems aim at providing advanced active help to the users of complex software systems. The performance of these help systems can be considerably improved if they are supplied with plan recognition and plan generation capabilities. Observing a user and recognizing his goals enables the system to help by taking into account the current state of the system as well as the user's level of education and current behavior. Moreover, if a planning capability is available user-specific support can be given by proposing appropriate plans which exactly is what the PHI system aims to achieve.

Cross-Talk Modes

One of the main characteristics of PHI consists in the close mutual cooperation between the planner and the plan recognizer. There are two cross-talk modes : In the first cross-talk mode the plan recognizer is able to determine the most likely plan a user follows by carrying out appropriate ``instantiations'' on valid plan hypotheses. Thus, services like semantic plan completion can be offered at any time during the observation process. The second cross-talk mode is devoted to providing the user with optimal plans whenever suboptimal behavior has been recognized or aid has explicitly been sought.

Planning and Plan Reuse

The generation of plans is based on standard assumptions concerning goals that typically occur or are specific to a certain user. Abstract plans are generated from these formal plan specifications. In doing so, the planner not only performs planning from first principles but is able to reuse already existing plans which are stored in a library (planning from second principles).

Plan Recognition

Plans serve as hypotheses in the recognition process. Taking abstract plans instead of concrete ones keeps the hypothesis space of manageable size. The plan hypotheses are passed to the recognition component where they are provided with numerical values which reflect the probabilities of their being confirmed by the subsequent observations. These a priori probabilities mirror a specific user's behavior, and are taken from the user model. Having observed the user's actions step by step the plan recognizer consequently tries to confirm the plan hypotheses by proving that the action sequence observed up to now is an admissible ``instance''. Hypotheses which are not confirmed are rejected and with that the probability distribution of the hypothesis space changes dynamically.

Abstract Plans

Abstract plans are those which represent a variety of ``concrete'' observable action sequences by admitting several degrees of freedom like variables (abstracting from the objects involved), abstract commands (abstracting from the names of actions which have the same effects), or temporal abstraction (abstracting from the point in time at which an action occurs).

The Planning Logic

The system is completely logic-based. It requires a proper axiomatization of the basic commands of the application system and certain domain constraints. The logic LLP which we have developed for that purpose combines features of both traditional programming and temporal logics. The plan generation and recognition components are special purpose inference procedures. Plan generation is done deductively using a sequent calculus for LLP, whereas plan recognition follows an abductive principle.

The Prototype

The application domain is a subset of the operating system UNIX, namely its mail system , where commands like type, delete, or manipulate objects like messages or mailboxes. Realizing plan generation and recognition in a strictly logic-based way nevertheless does not cause any considerable inefficiencies, even for ``real'' plans. This suggests evaluating the PHI approach even for richer application domains. A prototype of the PHI system has been implemented in SICSTUS PROLOG on a SUN SPARC computer.

The IGLOO-System

The system IGLOO is an interactive, graphic supported proof development environment for sequent calculi. It was implemented as a master thesis' work within the PHI project. It provides a valuable support for the development of proof guiding tactics. These tactics are necessary for an efficient deductive planner generating plans by proving plan specifications (LLP formulas specifying the goals to be reached by a plan) constructively.

Acknowledgement: The PHI Projekt was supported by the German Ministry for Research and Technology (BMFT) under contract ITW 90008