Verification of Numeric Planning Problems Through Domain Dynamic Consistency

Enrico Scala, Thomas L. McCluskey, Mauro Vallati

Abstract: Verification of the development of complex problem models is an open problem in real-world applications of automated planning. To facilitate the verification task, this paper introduces the notion of Domain Dynamic Consistency for planning problems expressed in PDDL. This notion is aimed at signalling suspicious inputs arising at the intersection between the abstract description of the model and its concrete instantiation. Together with the notion we present an approximation based approach that is devoted to automatically solve the problem of deciding when a PDDL numeric planning problem is not Domain Dynamic Consistent. The paper terminates with an example of application of this notion and its related technique within a Urban Traffic Control scenario.