Hierarchical Domain Knowledge Research Articles

Refining HTN Methods via Task Insertion with Preferences

Hierarchical Task Network (HTN) planning is showing its power in real-world planning. Although domain experts have partial hierarchical domain knowledge, it is time-consuming to specify all HTN methods, leaving them incomplete. On the other hand, traditional HTN learning approaches focus only on declarative goals, omitting the hierarchical domain knowledge. In this paper, we propose a novel learning framework to refine HTN methods via task insertion with completely preserving the original methods. As it is difficult to identify incomplete methods without designating declarative goals for compound tasks, we introduce the notion of prioritized preference to capture the incompleteness possibility of methods. Specifically, the framework first computes the preferred completion profile w.r.t. the prioritized preference to refine the incomplete methods. Then it finds the minimal set of refined methods via a method substitution operation. Experimental analysis demonstrates that our approach is effective, especially in solving new HTN planning instances.

HTN-like solutions for classical planning problems: An application to BDI agent systems

In this paper we explore the question of what characterises a desirable plan of action and how such a plan could be computed, in the context of systems that already possess a certain amount of hierarchical domain knowledge. In contrast to past work in this setting, which focuses on generating low-level plans, losing much of the domain knowledge inherent in such systems, we argue that plans ought to be HTN-like or abstract, i.e., re-use and respect the user-supplied know-how in the underlying domain. In doing so, we recognise an intrinsic tension between striving for abstract plans but ensuring that unnecessary actions, not linked to the specific goal to be achieved, are avoided. We explore this tension by characterising the set of “ideal” abstract plans that are non-redundant but maximally abstract, and then develop a more limited yet feasible account in which a given (arbitrary) abstract plan is “specialised” into one such non-redundant plan that is as abstract as possible. We present an algorithm that can compute such specialisations, and analyse the theoretical properties of our proposal.

Landmarks in Hybrid Planning

Although planning techniques achieved a significant progress during recent years, solving many planning problem still difficult even for modern planners. In this paper, we will adopt landmark concept to hybrid planning setting - a method that combines reasoning about procedural knowledge and causalities. Landmarks are a well-known concept in the realm of classical planning. Recently, they have been adapted to hierarchical approaches. Such landmarks can be extracted in a pre-processing step from a declarative hierarchical planning domain and problem description. It was shown how this technique allows for a considerable reduction of the search space by eliminating futile plan development options before the actual planning. Therefore, we will present a new approach to integrate landmark pre-processing technique in the context of hierarchical planning with landmark technique in the classical planning. This integration allows to incorporate the ability of using extracted landmark tasks from hierarchical domain knowledge in the form of HTN and using landmark literals from classical planning. To this end, we will construct a transformation technique to transform the hybrid planning domain into a classical domain model. The methodologies in this paper have been implemented successfully, and we will present some experimental results that give evidence for the consid- erable performance increase gained through planning system.

RECOGNIZING OVERLAPPED HUMAN ACTIVITIES FROM A SEQUENCE OF PRIMITIVE ACTIONS VIA DELETED INTERPOLATION

The high-level recognition of human activity requires a priori hierarchical domain knowledge as well as a means of reasoning based on that knowledge. Based on insights from perceptual psychology, the problem of human action recognition is approached on the understanding that activities are hierarchical, temporally constrained and at times temporally overlapped. A hierarchical Bayesian network (HBN) based on a stochastic context-free grammar (SCFG) is implemented to address the hierarchical nature of human activity recognition. Then it is shown how the HBN is applied to different substrings in a sequence of primitive action symbols via deleted interpolation (DI) to recognize temporally overlapped activities. Results from the analysis of action sequences based on video surveillance data show the validity of the approach.