Satis ability Planning with Causal Theories

A previous paper introduced the nonmono-tonic formalism of causal theories, along with a general method for representing action domains in it. Here we show that causal action theories provide a basis for eeective automated planning. To this end, we deene several properties plans may have, such as executability, determinism, and validity. We then identify a class of causal action theories for which (i) there is a concise translation into classical logic, and (ii) the models of the resulting classical theories correspond to valid plans. These results enable satiss-ability planning (in the sense of Kautz and Selman) on the basis of action formalizations that include indirect eeects of actions (ramii-cations), implied action preconditions (quali-cations), concurrent actions, and other features of action domains. Causal theories representing the large blocks world and logistics planning problems from Kautz and Sel-man, 1996] are solved comparatively quickly, demonstrating the eeectiveness of our approach .