Description of Prodigy
Description of Prodigy
A diagram of the Prodigy architecture
The Prodigy, as embodied in Prodigy, is a modular architecture that
stores its knowledge symbolically in
a form of
first-order predicate logic (FOPL)
which is called the Prodigy Description Language (PDL).
This language includes:
- Ground Atomic formulas: used in database to express current state of world,
- Operators: used to represent what actions effect which changes,
- Inference Rules: used to deduce added information about a state,
- Control Rules: used to guide search, and
- Meta-knowledge axioms: used to achieve
self-reflection
All knowledge is therefore declarative,
freely-inspectable, and uniformly
represented and
accessed by all modules.
Prodigy's planning approach is
STRIPS-like in that it uses
Means-Ends Analysis and control rules to
choose:
- which node to expand,
- which goal to work on,
- which operator to apply, and,
- which instatiation of an operator to select.
MEA and the control rules give Prodigy
focused behavior.
By default, Prodigy generates new goals in a depth-first manner.
One of the goals of Prodigy was to be a
test-bed for learning mechanisms.
As such, Prodigy supports many learning
methods.
Among these are:
Return to the top of this architecture.
Current Location: Prodigy - Architectural Description