Distributed Supervisory Control of Large Plants (DISC)
Supervisory control is a formal approach for the control of discrete event systems that aims to solve logical problems of safety, resource allocation, liveness, and fault diagnosis that can be encountered in all systems with a high degree of automation. It provides a conceptual framework for developing methods and tools for system design.
An open issue is the application of this methodology to those control problems that arise in networked embedded systems. These distributed plants are compromised by several local agents that take concurrently decisions, based on information that may be local or received from neighbouring agents; they require scalable and self-organising platforms for advanced computing and control.
An important feature of this type of processes is the possibility of studying them at an appropriate level of abstraction where the resulting model is a purely discrete event one. The evolution is guided by the occurrence of asynchronous events, as opposed to other real-time models where the event occurrence is time-triggered.
We plan to use several techniques to reduce the computational complexity that is one of the major obstacles to the technology transfer of supervisory control methodologies to distributed plants. These techniques are: modularity in the moddeling and control design phases; coordinating control; modular state identification and modular fault detection based on the design of decentralized observers.