On Conflicts and Strategies in QBF
Nikolaj Bjørner, Mikoláš Janota, William Klieber · EPiC series in computing · 2018
QBF solving techniques can be divided into the DPLL-based and expansion-based. In this paper we look at how these two techniques can be combined while using strategy extraction as a means of interaction between the two. Once propagation derives a conflict for one of the players, we analyse the proof of such and devise a strategy for the opponent. Subsequently, this strategy is used to constrain the losing player. The implemented prototype shows feasibility of the approach. A number of avenues for future research can be envisioned. Can strategies be employed in a different fashion? Can better strategy be constructed? Do the presented techniques generalize beyond QBF?