Means-End Analysis in Prodigy
Means-End Analysis in Prodigy
Prodigy's problem solver uses a means-ends
algorithm to solve problems. Differences are determined by comparing
the current state and the goal state. Unlike most means-ends
algorithms, Prodigy may have several simultaneous goals to consider.
The system first determines which goal to work toward and then
generates differences for an individual goal once this determination
has been made. These differences may be readily computed since all
knowledge is expressed in PDL, which is both
homogeneous and penetrable. Once differences have been
computed for the current goal, operators are proposed based on their
ability to reduce the differences. Since Prodigy uses STRIPS-like operators, this step corresponds to
scanning the current operators' add-lists to determine if an operator
asserts any predicates which, after binding, would result in the
removal of a difference (the assertion of a goal state conjunct).
Selection among the operators is mediated by control rules. In the absence of control
rules, in accordance with Prodigy's casual
commitment strategy, the mediation between operators defaults to a
depth-first consideration of each.
Return to the top of this architecture.
Go to a discussion of this property
for multiple architectures.
Current Location: Prodigy - Properties - Means-End Analysis