Modular Action Language ${\cal ALM}$

Abstract

The field of knowledge representation experienced substantial progress in representing dynamic domains during the last decades. Action languages were created to describe transition diagrams in a mathematically accurate and concise manner. Solutions for the frame, ramification and qualification problems were discovered. We now have a good understanding of how to describe effects of actions and action executability conditions. However, the issue of describing objects of dynamic domains, including actions and fluents, remains almost unaddressed. The traditional approach, in which such objects are represented as constants or terms, doesn’t allow for an elaboration tolerant and scalable represention of action properties. In previous action languages it is impossible to describe objects of the domain in terms of other, already defined, objects. This and similar features can be achieved by adding modularity to action languages. This would enable programmers to create libraries of knowledge about dynamic domains.