This talk will present the work of my PhD thesis on fault tolerant architectures for distributed systems. Spatially distributed (or large scale) systems are composed by nodes distantly located and communicating through a network to reach a global goal. Each node, anyway, has its local objective. For this reason the supervision and control methods I studied for distributed systems reflect their decentralized nature. In particular their structural properties of scalability and modularity have been considered. Moreover a hierarchical architecture helps in managing communication issues.
In the first part of my thesis distributed systems are modelled as discrete event systems, because only the communication policy between the nodes and the different discrete states of the system are considered. But nodes have in general continuous dynamics; hence a deeper study involves the use of hybrid systems as a modelling framework. The talk will then address some open problems on diagnosis of hybrid systems and networks.
Finally the talk will introduce decentralized methods for control and estimation of distributed systems and some open issues on this topic.