COMMENT ⊗   VALID 00002 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	travel[e83,jmc]		The travel problem in FOL and Prolog
C00007 ENDMK
C⊗;
travel[e83,jmc]		The travel problem in FOL and Prolog

	The object is to use the problem of planning a trip as
a focus for formalization.  It offers the simplification that it
is an essentially symbolic problem, i.e. there are not really
any extended objects or scenes to recognize.  Ultimately we would
like our formalization to take into account the following facts.

1. Situations change, so facts must either change their truth
values or must be parametrized by time or situation.

2. Part of travelling is getting information.

3. Obstacles can arise and must be overcome.

4. The trip is planned at various levels of detail.  Suppose we
adopt the experienced traveler's habit of postponing all decisions
that can be postponed.  Therefore, we prove that booking the flight
is the only action that has to be taken immediately and also the
only decision that has to be taken immediately.  Defaults tell us
that planning the trip to the airport can be postponed.
Indeed the initial plan to attend IJCAI is to plan to go to Karlsruhe.
Even deciding on a flight is postponable.

5. Questionable whether we should try to include this in the travel
problem, but here goes.  A fact that is true in a context
defaults to true in a wider context.

Features of travel to be taken into account.

1. While I'm there I might as well accept Stoyan's invitation
to go to Erlangen.

2. airlines, passports and visas, strikes and other stoppages,
requirements for money, flight cancellations.

3. Reifying concepts gives no difficulty except when a propositional
attitude must be applied to an existential statement.  Universals
give less trouble.

Perhaps both ordinary backtracking and dependency directed backtracking
correspond to features of common sense problem solving.  To put a question:
Is ordinary backtracking dominated by dependency directed backtracking?
Can dependency directed backtracking be improved by a killer heuristic?
Namely when we are achieving a conjunction of goals, and a goal fails,
try to put the goal that failed at the head of the list.  This will
sometimes be excluded by the constraints.  The analogy with chess
is that in chess, we are trying to make a move that will not be
refuted by any opponents move, so anticipating these moves is a
conjunctive set of goals.

The blocks world may serve as a model for a lot of other problems.
It would be an elaborated blocks world allowing more complex conditions
for placing a block.  Perhaps we represent partial information about
the other problem by a totally defined blocks world.  Then we can
try out hypotheses. - rather vague.