Section: Application Domains

Opacity, Supervision, and Petri Nets

Our activities on components emerged from a larger basis of competences developed in the past of S4 on supervisory control and Petri net synthesis. Components and their interfaces are intimately tied to supervisory control, and Petri net synthesis is a possible approach to controller synthesis. In the last four years, we have carried on work on both themes, but refocussed our research on fresh topics. A major contribution has been to study supervisory control for secrecy objectives, with promising results. Another contribution has been to study supervisory control for finite abstractions of services. The fusion of both topics, that would increase the interest of the results for Web applications, is not yet done. A different topic that we continued to investigate is the synthesis of distributed controllers based on the synthesis of distributed Petri nets. Our progress on this difficult topic is limited, but we feel we should pursue the effort.

Opacity is an abstract property that includes non-interference and that can cover confidentiality, authenticity and many other specific security concepts. Our project-team has inaugurated research on supervisory control of discrete event systems for opacity, which became soon a theme of cooperation with project-team Vertecs and subsequently attracted concurrent researches at Wayne State U., Kyoto Inst. of Tech., and U. Illinois. We have some advance over these concurrent teams.

The rest of our work on supervision focusses on minimizing communication between decentralized controllers, on asynchronous and distributed control, and on the enforcement of modal specifications. Decreasing communication between decentralized controllers was studied at Michigan U. but we could further show that minimizing communication reduces to a classical optimization problem. As regards asynchronously communicating control, the only current attempts we are aware of are those of project-teams S4 and Vertecs. As regards supervisory control w.r.t. modal specifications, the closest work is Lohmann and Wolf's synthesis of communication partners for Web services.

The approach which we propose towards distributed control relies upon the synthesis of distributed Petri nets. We have been leaders for fifteen years on the synthesis of P/T nets, on a par with the Petrify team focussed on Elementary (or safe) net synthesis. The algorithms which we have defined have been reused or adapted by many other researchers in Europe, in the US, and in China, to respond to three types of problems: controller synthesis, process mining, and concise representation of services. We are currently writing a book covering all aspects of the theory and applications of Petri net synthesis. We also pursue research on structure theory of Petri nets, in cooperation with U. Oldenburg, with focus set recently on non-interference.