A Graph-Transformation Modelling Framework for Supervisory Control

Abstract

Formal design methodologies have the potential to accelerate the development and increase the reliability of supervisory controllers designed within industry. One promising design framework which has been shown to do so is known as supervisory control synthesis (SCS).

In SCS, instead of manually designing the supervisory controller itself, one designs models of the uncontrolled system and its control requirements. These models are then provided as input to a special synthesis algorithm which uses them to automatically generate a model of the supervisory controller. This outputted model is guaranteed to be correct as long as the models of the uncontrolled system and its control requirements are valid. This accelerates development by removing the need to verify and rectify the model of the supervisory controller. Instead, only the models of the uncontrolled system and its requirements must be validated.

To address problems of scale, SCS can be applied in modular fashion, and implemented in hierarchical and decentralized architectures.

Despite the large body of research con rming the bene ts of integrating SCS within the development process of supervisory controllers, it has still not yet found widespread application within industry. In the author's opinion, this is partly attributed to the non-user-friendly nature of the automaton-based modelling framework used create the models of the uncontrolled system (and control requirements in even-based SCS). It is believed that in order for SCS to become more accessible to a wider range of non experts, modelling within SCS must be made more intuitive and user-friendly.

To improve the usability of SCS, this work illustrates how a graph transformation-based modelling approach can be employed to generate the automaton models required for supervisory control synthesis. Furthermore, it is demonstrated how models of the speci cation can be intuitively represented within our proposed modelling framework for both event- and state-based supervisory control synthesis. Lastly, this thesis assesses the relative advantages brought about by the proposed graph transformation-based modelling framework over the conventional automaton based modelling approach.