This paper postulates a rigorous method for the construction of classical planning domain models. We describe, with the help of a non-trivial example, a tool supported method for encoding such models. The method results in an `object-centred' specification of the domain that lifts the representation from the level of the literal to the level of the object. Thus, for example, operators are defined in terms of how they change the state of objects, and planning states are defined as amalgams of the objects' states. The method features two classes of tools: for initial capture and validation of the domain model; and for operationalising the domain model (a process we call compilation) for later planning. Here we focus on compilation tools used to generate macros and goal orders to be utilised at plan generation time. We describe them in depth, and evaluate empirically their combined benefits in plan-generation speed-up. The method's main benefit is in helping the modeller to pro...

Hierarchical task network and action-based planning approaches have traditionally been studied separately. In many domains, human expertise in the form of hierarchical reduction schemas exists, but is incomplete. In such domains, hybrid approaches that use both HTN and action-based planning techniques are needed. In this paper, we extend our previous work on refinement planning to include hierarchical planning. Specifically, we provide a generalized plan-space refinement that is capable of handling non-primitive actions. The generalization provides a principled way of handling partially hierarchical domains, while preserving systematicity, and respecting the user-intent inherent in the reduction schemas. Our general account also puts into perspective the many surface differences between the HTN and action-based planners, and could support the transfer of progress between HTN and action-based planning approaches. 1 Introduction Traditionally, classical planning probl...

Early work on the NONLIN and O-Plan projects indicated a need for a defined methodology which would guide users performing various roles in the acquisition and analysis of domain requirements for planning. This work included links to a requirement analysis methodology, CORE (COntrolled Requirements Expression) , tool support via an intelligent assistant as part of the Task Formalism (TF) Workstation and an initial collection of guidelines and checklists to aid in using the TF domain description language. This paper describes work underway to follow-on from this past research and to infuse it with knowledge gained from recent research related to planning domain development, knowledge modelling, design rationale and ontological and requirements engineering. Introduction The activities involved in discovering, engineering, documenting, and maintaining a set of domain constructs for most AI planning-based projects can be considered ad hoc and disorganised, at best. The current sources for...

Dynamic Flexible Constraint Satisfaction and its Application to AI Planning Constraint satisfaction is a fundamental Arti cial Intelligence technique for knowledge representation and inference. It has, however, become clear that the original formulation of a static constraint satisfaction problem (CSP) with hard, imperative constraints is insucient to model many real problems. Recent work has addressed these shortcomings in the form of two separate extensions known as dynamic CSP and exible CSP respectively. Little has yet been done to combine dynamic and exible CSP in order to bring to bear the bene ts of both in solving more complex problems.

State space and plan space planning approaches have traditionally been seen as fundamentally different and competing approaches to domain-independent planning. We present a plan representation and a generalized algorithm template, called UCP, for unifying these classical planning approaches within a single framework. UCP models planning as a process of refining a partial plan. The alternative approaches to planning are cast as complementary refinement strategies operating on the same partial plan representation.

Currently, few tools are available for assisting developers with debugging intelligent systems. Because these systems rely heavily on context dependent knowledge and sometimes stochastic decision making, replicating problematic performance may be di cult. Consequently, we adopt a statistical approach to modeling behavior as the basis for identifying potential causes of failure. This paper describes an algorithm for constructing state transition models of system behavior from execution traces. The algorithm is the latestinafamily of statistics based algorithms for modelling system execution called Dependency Detection. We present preliminary accuracy results for the algorithm on synthetically generated data and an example of its use in debugging a neural network controller for a race car simulator. 1

Evaluation and debugging of AI systems require coherent views of program performance and behavior. We have developed a family of methods, called Dependency Detection, for analyzing execution traces for small patterns. Unfortunately, these methods provide only a local view of program behavior. The approach described in this paper integrates two methods, dependency detection [10] and CHAID-based analysis [3], to produce an abstract model of system behavior: a transition diagram of merged states. We present the algorithm and demonstrate it on synthetic examples and data from two AI planning and control systems. The models produced by the algorithm summarize sequences and cycles evident in the synthesized models and highlight some key aspects of behavior in the two systems. We conclude by identifying some of the inadequacies of the current algorithm and suggesting enhancements.

In this paper, we introduce a domain modeling tool that supports users in the incremental and modular development of verified models of planning domains. It relies on a logic-based concept for systematic domain model construction that provides well-defined, safe operations for the union, extension, and refinement of already existing models. The system is equiped with a deductive component. It automatically performs the proofs necessary to guarantee both the consistency of single models and the safety of operations on models. By means of detailed examples, it is shown how the system has been used for the structured development of a model for a complex, safety-critical planning domain. 1 Introduction As soon as we aim at using planning systems in the context of realistic applications, the task of generating the underlying domain model becomes increasingly crucial. It is not only difficult to overlook the great amount of object types, relations, and actions involved when specifying such...

Process engineering involves a search for new models of organising work. This synthesis task can become quite difficult and time-consuming as the amount of detail required and interactions between activities increases. Domain independent AI planning offers some promising techniques and representations to assist in this effort. One of the major impediments to transferring this technology to applied, real-world settings is the difficulty encountered in building the domain model which is used in the automated generation of these plans. Competence, as well as good tools, is necessary to carry out this task. A plan domain methodology should be available which provides structured organisational development activities. Users need to know what tasks they have to perform: for each step, information must be available about what input will be needed, and what output will be required, what is to be done and how it can be done well. This paper presents the Common Process Methodology (CPM) which aim...

This paper describes a multistrategy learning approach for automatic acquisition of planning operators. The two strategies are: (i) learning operators by observing expert solution traces, and (ii) refining operators through practice in a learning-by-doing paradigm. During observation, OBSERVER uses the knowledge that is naturally observable when experts solve problems, without the need of explicit instruction or interrogation. During practice, OBSERVER generates its own learning opportunities by solving practice problems. The inputs to our learning system are: the description language for the domain, experts' problem solving traces, and practice problems to allow learningby -doing operator refinement. Given these inputs, our system automatically acquires the preconditions and effects (including conditional effects and preconditions) of the operators. Our approach has been fully implemented in a system called OBSERVER on top of a non-linear planner PRODIGY.We present empirical results t...