Diagnosing plan execution discrepancies 
		in a logic-based action framework

	  Thomas Eiter, Esra Erdem, and Wolfgang Faber 
		Vienna University of Technology
	

This paper introduces a logic-based framework for monitoring plan
execution relative to a given set of trajectories.  According to this
framework, a monitoring agent can tell when things go wrong by checking
the compatibility of the plan execution with the given trajectories,
considering the current state information.  She finds an explanation for
these detected discrepancies by examining the given trajectories against
possible evolutions of the current state from the initial state.  Such an
explanation provides information about possible points of failure that may
be useful in two ways. First, points of failure can be used to determine
some checkpoints (specifying when to check for discrepancies), and this
may be useful if the plan is executed many times. Second, they can be used
for recovering from discrepancies. For instance, the agent can undo the
plan execution until a point of failure, from which the remainder of the
plan can be (re)executed, thus avoiding an expensive (re)planni ng step.  
In this paper, we present formal specifications for discrepancies and
explanations for them in a general action representation framework, which
can accommodate nondeterminism, concurrency, and dynamic worlds.  We
describe two ways of finding explanations for detected discrepancies, and
study their properties. Moreover, we analyze the computational complexity
of detecting discrepancies and finding explanations for them, and describe
how these computational problems can be solved.