Resumen de Knowledge engineering techniques for the translation of process models into temporal hierarchical planning and scheduling domains
Arturo González Ferrer 
Artificial Intelligent Planning & Scheduling technology (AI P&S) has confronted, from its very beginning, the absolute necessity of improving the support for using and integrating traditional software engineering techniques, or at least, to get closer to the users of the application domain where AI P&S components want to be integrated. This is due to the fact that the development of components based on AI P&S, has traditionally used ad-hoc languages and techniques developed for the modeling of real complex scenarios, separated from traditional software development methods. Furthermore, in these scenarios it is typically needed to manage information such as actions, states and constraints about time and resources, using a representation that is nor trivial neither usual in other areas of Computer Science.
It is evident that, in order to achieve this aim, it is needed to develop Knowledge Engineering techniques that allow to accomplish the stages of expert knowledge acquisition and representation, in a similar way to how it is carried out in other areas of Computer Science, either by means of traditional techniques for software design, or by means of methods and languages used in the specific application field. This aims to simplify the task of engineers and the interaction with experts when modeling real problems, allowing to improve the reuse, and facilitating at the same time the maintenance of components based on AI P&S techniques. In this sense, this dissertation is aligned with the above mentioned goals, and aims to develop Knowledge Engineering techniques to automatically carry out the translation of process models, in order to express them in terms of a representation based on the HTN (Hierarchical Task Network) planning paradigm.
Concretely, the dissertation is focused on the study of two specific application domains: business process models expressed by means of the BPMN notation, and clinical processes expressed by means of Clinical Practice Guidelines (CPGs). In the former, it is studied how to carry out the translation of control-flow patterns commonly found on such models, as well as the representation of resource constraints. On the latter, the previous work is extended, in order to translate complex temporal constraints (e.g. temporal annotations, synchronizations and cycles) that are fundamental to represent process models in the clinical domain.
The results of this research allow, on the one hand, to improve the capability to define planning domains models in a simple way, starting from standard notations used for the process modeling in the mentioned application fields. On the other hand, they introduce features for decision making support in such application domains. This last contribution allows to solve problems difficult to solve so far in both domains, such as the automatic generation of working plans in organizative processes, or the automatic generation of patient-tailored treatment plans, adding extra value to AI P&S technology.
