QBF reasoning on real-world instances

  • Authors:

  • Enrico Giunchiglia;Massimo Narizzano;Armando Tacchella

  • Affiliations:

  • DIST, Università di Genova, Genova, Italy;DIST, Università di Genova, Genova, Italy;DIST, Università di Genova, Genova, Italy

  • Venue:
  • SAT'04 Proceedings of the 7th international conference on Theory and Applications of Satisfiability Testing
  • Year:
  • 2004

Quantified Score

Hi-index 0.00

Visualization

Abstract

During the recent years, the development of tools for deciding Quantified Boolean Formulas (QBFs) satisfiability has been accompanied by a steady supply of real-world instances, i.e., QBFs originated by translations from application domains such as formal verification and planning. QBFs from these domains showed to be challenging for current state-of-the-art QBF solvers, and, in order to tackle them, several techniques and even specialized solvers have been proposed. Among these techniques, there are (i) efficient detection and propagation of unit and monotone literals, (ii) branching heuristics that leverages the information extracted during the learning phase, and (iii) look-back techniques based on learning. In this paper we discuss their implementation in our state-of-the-art solver QuBE, pointing out the non trivial issues that arised in the process. We show that all the techniques positively contribute to QuBE performances on average. In particular, we show that monotone literal fixing is the most important technique in order to improve capacity, followed by learning and the heuristics. The situation is reversed if we consider productivity. These and other observations are detailed in the body of the paper. For our analysis, we consider the formal verification and planning benchmarks from the 2003 QBF evaluation.