Augmenting JSHOP2 planning with OWL-DL

Abstract

The paper describes an approach for combining a reasoner based on description logic (concretely, OWL-DL) and a planner, in this case an off-the-shelf HTN planner (concretely, JSHOP2). The domain representation is formulated in the DL regime, which can also used for logical reasoning about state. Domain and problem descriptions are then generated automatically from the DL representation as needed, plans are generated on the planner's side, and transformed back into the DL representation. A working example is given for a simulated robot navigation domain. In particular, we show that the automatic planning problem generation leads to skipping over parts of the domain description deduced to be irrelevant for solving the concrete planning problem at hand, thereby reducing substantially the size of the problem description for the planner and speeding up the planning process.