We present a new approach for preprocessing Quantified Boolean Formulas (QBF) in conjunctive normal form (CNF) by expanding a selection of universally quantified variables with bounded expansion costs. We describe a suitable selection strategy which exploits locality of universals and combines cost estimates with goal orientation by taking into account unit literals which might be obtained. Furthermore, we investigate how Q-resolution can be integrated into this method. In particular, resolution is applied specifically to reduce the amount of copying necessary for universal expansion. Experimental results demonstrate that our preprocessing can successfully improve the performance of state-of-the-art QBF solvers on well-known problems from the QBFLIB collection.
展开▼
