Using Finite-State Automata to Model and Solve Planning Problems

In the recent years, the representation using the state variables, i.e., SAS+ planning, has become one of the most popular formalisms for specifying the planning problems (and domains). Still, the great majority of the planning tasks, especially from the past planning competitions, are defined using the classical STRIPS-like formalism, and so an efficient transformation technique is required in order to allow performance comparison of individual planners. In the paper we present the new formalism that is based on finitestate automata (FSA), and that represents an alternative to the SAS (state variables), providing thus a novel platform for development of the modern planners. Finally, we also propose conversion procedures between the individual planning problem representations, together with the basics of automated and semi-automated techniques that employ the new formalism and aim towards the extraction of the state variables representation from the classical representation.