Dealing with the unexpected obstacle

One makes a plan, but part way through its execution an
unexpected obstacle appears.  The plan is revised, the
obstacle is overcome, and the goal of the plan is realized.
Humans do this all the time, but the usual AI forms of
planning don't provide for it in a sufficiently general way.

Here's our general idea for dealing with it.

The original plan is generated from the facts of the
situation and goal combined with information from
the general common sense database.  The process of
generating the plan involves  circumscription.
During the execution of the plan an obstacle appears.
To the reasoning mechanism the obstacle takes the form
of an unexpected fact.  When the new fact is added
to the set of facts originally taken into account,
the circumscriptions no longer give the original
plan but rather a new plan that overcomes the
obstacle.

The problem is the details of this.  We need a good
example.