Compositional Approaches in Supervisory Control with Application to Automatic Generation of Robot Interlocking Policies

Doctoral thesis, 2006

The work presented in this thesis concerns verification and synthesis in the Ramadge and Wonham supervisory control framework. Supervisory control constitutes a formal framework for the design of supervisors for discrete event systems. These systems usually model high level descriptions of logical behaviours in applications such as flexible manufacturing processes, chemical batch processing systems and communication systems. The supervisory control framework has the potential to solve many safety and flexibility issues in such systems. Unfortunately, the analysis of discrete event systems involves an intrinsic difficulty known as the state-space explosion problem---a combinatorial explosion that soon occurs when problems of real-world complexity are analysed. The state-space explosion problem has given rise to much research and in the last decades many ingenious approaches to solving the problem have been presented. However, most of these approaches have in common that they can only be applied to special classes of supervisory control problems or that they only give partial solutions. Therefore, in a new attempt to overcome the state-space explosion problem for general supervisory control problems, this thesis develops compositional methods for verification and synthesis in the supervisory control framework. Compositional methods exploit the inherent modularity of discrete event models by using abstractions to incrementally hide already analysed behaviour of the system. A major part of the thesis concerns developing a methodology for calculating these abstractions and for applying the compositional approach to complex problems. Furthermore, a very important part of this work is the implementation of the compositional methods in a software tool for supervisory control. Experimental results from this implementation are also presented in the thesis. As a matter of fact, earlier work on computational complexity have shown that it is impossible to solve the state-space explosion problem efficiently for general problems. Even so, the results presented in this thesis show that ``large'' supervisory control problems found in the litterature typically have enough structure in them to be solved efficiently by compositional methods. Another contribution of this thesis is the development of a method to automatically generate models of the necessary interlocking requirements with respect to robot collisions in industrial robot cells. The generated models are suitable for supervisor synthesis in the supervisory control framework as well as for work-cycle optimisation. Automatic model generation and synthesis are important factors for shortening the development time and improving the flexibility for industrial robot cells.