Causal action theories and satisfiability planning.

摘要

his dissertation addresses the problem of representing and reasoning about commonsense knowledge of action domains. Until recently, most such work has suppressed the notion of causality, despite its central role in everyday talking and reasoning about actions. There is good reason for this. In general, causality is a difficult notion, both philosophically and mathematically. Nonetheless, it turns out that action representations can be made not only more expressive but also mathematically simpler by representing causality more explicitly. The key is to formalize only a relatively simple kind of causal knowledge: knowledge of the conditions under which facts are caused. In the first part of the dissertation we do this using inference rules and rule-based nonmonotonic formalisms. As we show, an inference rule