Information Technology Reference
In-Depth Information
alive ( t ) ;
: fleeing ( t )
alive ( t ) ;
fleeing ( t )
startle(t)
: alive ( t ) ;
: fleeing ( t )
: alive ( t ) ;
fleeing ( t )
6; f (2 : 16) g
startle(t)
?
?
Figure 2.12. Startling the turkey is possible only if the animal is alive. This im-
plicit precondition is encountered by failing to transform the preliminary successor
f: alive ( t ) ; fleeing ( t ) g into an acceptable state via the only available causal re-
lationship (2.16).
be alive if we want to startle it with success; see Fig. 2.12. We see that our
constraint may also give rise to implicit qualications.
This completes our formal account of the Ramication Problem. Let us
summarize: A ramication domain contains a number of state constraints, of
which some are steady, that give rise to indirect eects of actions. With the
aid of general information as to possible influences among the fluents, the
potential indirect eects, formalized as so-called causal relationships, can be
automatically extracted from the constraints. This helps to distinguish the
correct eects from both unmotivated changes and implicit qualications.
Causal relationships are successively applied to preliminary successor states
until all indirect eects have been accounted for; in the course of this process
the so-called coupled eects, deriving from steady state constraints, are to
be generated prior to any so-called triggered eect.
2.9 A Fluent Calculus Axiomatization
In this section, we develop an axiomatization of our action theory by means
of standard logic. The major motivation for so doing is that action theo-
ries, as they stand, are less suited than general purpose logics when it comes
to automating reasoning. The reason is that entailment relations of action
theories are usually much more specialized and sophisticated because they
reflect complex notions such as time, change, and causality. Moreover, when-
ever an action theory is modied or extended in order to address additional
Search WWH ::




Custom Search