State-Based Planning with Numerical Knowledge

The drawback of the STRIPS encoding to support numbers in planning has motivated the extension to PDDL language the PDDL2.1 version to allow numerical state variables. Many planning systems as FF-Metric and SAPA treat numbers in planning, but only resources and time are represented as numbers in their planning problems. In contrast, some real world problems require an enormous numerical representation. Therefore, we present in this paper a new planning framework allowing the application of arithmetic functions for numerical knowledge update. We propose a new action representation to improve numerical conditions and effects, which allow the application of symbolical planning techniques over numerical knowledge. As an application of our planning framework, we show the implementation of a numerical version of the well-known FF planning system able to solve domains containing symbolical and/or numerical knowledge.