@device(postscript)
@libraryfile(Mathematics10)
@libraryfile(Accents)
@style(fontfamily=timesroman,fontscale=11)
@pagefooting(immediate, left "@c<Technical Report 1996>",
center "@c<Computer Science>",
right "@c<Carnegie Mellon>")
@heading(Learning Planning Operators by Observation and Practice)
@heading(CMU-CS-96-154)
@center(@b(Xuemei Wang))
@center(June 1996 - Ph.D. Thesis)
@center(FTP: Unavailable)
@blankspace(1)
@begin(text)

Acquiring and maintaining domain knowledge is a key bottleneck in 
applications of planning systems.  This thesis describes a machine
learning approach to automatic acquisition of planning operators.
Our approach is to learn planning operators by observing expert
solution traces and to refine operators through practice in a
learning-by-doing paradigm. During observation, our system used the
knowledge that is observable when experts solve problems, without the
need of explicit instruction or interrogation.  During practice, our
system generates its own learning opportunities by solving practice problems.

The inputs to our learning system are: the description language for the
domain, experts' problem solving traces, and practice problems to allow
learning-by-doing operator refinement.  The output is a set of operators,
each described by a list of variables, preconditions, and effects. The
operators are learned incrementally using an inductive algorithm.  During
practice, our system effectively generates plans using incomplete and
incorrect operators, repairs plans upon execution failures, and
integrates planning, learning, anad execution.

Our approach has been fully implemented and tested in a systems
called @c(Observer), which is built in the context of the PRODIGY4.0
nonlinear planner. Empirical results in a process planning domain and a
version of the 34 meter Deep Space Network antenna operation domain
demonstrate the validity of our approach.  These results show that our
system learns operators in these domains well enough to solve problems as
effectively as expert human-coded operators, and that learning by
observation and learning by practice both contribute significantly
to the learning process.

@blankspace(2line)
@begin(transparent,size=10)
@b(Keywords:@ )@c<Artificial intelligence, planning, problem solving,
PRODIGY, knowledge acquisition, machine learning, observation, practice>
@end(transparent)
@blankspace(1line)
@end(text)
@flushright(@b[(149 pages)])