Flexible Planning Knowledge Acquisition for Industrial Processes

The acquisition of planning knowledge for industrial processes has been shown to be particularly hard in terms of acquiring robust task knowledge from human experts. Furthermore, this knowledge, when acquired, is rather domain dependent and the acquisition effort is very seldom amortized over other applications. In this paper, we report on our work on flexibly representing a functional model of a chemical process in an Artificial Intelligence (AI) planning system. We show the generality and flexibility of our new planning operator representation. Operators are designed for general functions which can be defined for hierarchical levels of functional detail. We present how the planning domain can be used with little or no modification, depending on the level of abstraction, to other engineering systems. This planning domain is fully implemented in a domain-independent AI planning system.