Optimizing Planning Domains by Automatic Action Schema Splitting

As modeling details can have a large impact on planner performance, domain transformation has been a traditional subject of interest in the planning community not only between languages, but also within languages. Herein, we automate an intra-language transformation method that has as yet been applied only manually, and that has never been formally described: action schema splitting, which transforms an action schema with a big interface (many parameters) into several schemas with smaller interfaces, exponentially reducing the number of ground actions. We spell out this method, characterizing exactly the choice of splits preserving equivalence to the original schema. Making that choice involves a trade-off between interface size and plan length, which we explore by designing automatic domain optimization methods. Our experiments show that these methods can substantially improve performance on domains with large interfaces.