Hierarchical Planning through Operator and World Abstraction using ontology for home service robots
Dongyeop Kang, Ho-Jin Choi · International Conference on Advanced Communication Technology · 2009
In order to reduce the number of search space in planning, many abstraction techniques are introduced. For example, several hierarchical planners already provide significant reduction in search space. Hierarchical planning in real world, however, requires not only planning knowledge such as plan operators but domain knowledge used for describing world model. One of most important issue for Hierarchical Planning through Operator and World Abstraction (HPOWA) is to model an efficient, semantic, and flexible knowledge representation for both planning and domain knowledge. From the profound investigation of existing abstraction-based planning researches, we obtain several limitations of hierarchical planning for real world environments. Thus, we apply ontology to model knowledge representation for operator, world model and their interaction. Constructed robot ontology consists of action, object and context classes. The three layers of concepts link up operator knowledge with domain knowledge in world model. The concepts for each class are hierarchically represented for efficient hierarchical reasoning. The semantic relations among the domain concepts or sequential relations between plan operators are represented as relations and attributes of the ontology.