Expressiveness of ADL and Golog: Functions Make a Difference

The main focus in the area of action languages, such as GOLOG, was put on expressive power, while the development in the area of action planning was focused on efficient plan generation. An integration of GOLOG and planning languages would provide great advantages. A user could constrain a system’s behavior on a high level using GOLOG, while the actual low-level actions are planned by an efficient planning system. First endeavors have been made by Eyerich et al. by identifying a subset of the situation calculus (which is the basis of GOLOG) with the same expressiveness as the ADL fragment of PDDL. However, it was not proven that the identified restrictions define a maximum subset. The most severe restriction appears to be that functions are limited to constants. We will show that this restriction is indeed necessary in most cases.

• Type: Conference
• Year: 2007
• Where: AAAI
• Authors: Gabriele Röger, Bernhard Nebel
• Keywords: AAAI 2007, Actual Low-level Actions, Efficient Plan Generation, Efficient Planning System, Intelligent Agents