Hierarchical control reconfiguration for a class of hybrid systems

Active approaches to Fault-Tolerant Control seek to maintain safety and availability of a plant by reconfiguring its control system when a fault occurs. The fundamental problem in this approach is complexity: closing the diagnosis-reconfiguration-control loop online poses a major challenge. This is particularly hard for hybrid Systems, which are as yet not well understood.; In this dissertation a hierarchical approach is considered aimed at lowering the complexity of low-level control. The hierarchy consist of a supervisory controller that guides the system through a sequence of regions in the state space, and a configuration manager that selects the plant inputs for a low-level controller to use, and the low-level controller. The theory of invariant sets provides the necessary tools for the reachability calculations, which are performed at design time. Using model-predictive control as a case in point, it is shown that the hierarchical architecture facilitates the reduction of the complexity of the low-level control in three ways: reducing the number of discrete modes of the plant that need to be considered, reducing the prediction horizon needed for control and reducing the number of plant inputs manipulated by the controller. A three-tank system is used to illustrate the approach. The system model which is considered is a constrained discrete-time piecewise-affine system with additive state disturbance.; The architecture requires enumeration of configurations but not enumeration of faults. A feature of the hierarchical architecture, is that controller configurations can be designed for each supervisory control mode separately and control laws can be designed separately for each configuration. The significance from a practical point of view, is that design of fault-accommodation logic can be simplified by adopting the hierarchical approach even if it is designed by conventional simulation and testing methods, rather than formal verification.