A bank of reconfigurable LQG controllers for linear systems subjected to failures

Abstract

An approach for controller reconfiguration is presented. The starting point in the analysis is a sufficiently accurate continuous linear time-invariant (LTI) model of the nominal system. Based on a bank of reconfigurable LQG controllers, each designed for a particular combination of total faults, the reconfiguration consists of two operation modes. In the first mode a switching is invoked towards one of the pre-designed LQG controllers on the basis of the information about only the combination of total faults that is in effect. In the second mode, which is activated in cases of partial and component faults, a dynamic correction procedure is initiated which tries to reconfigure the currently active controller in such a way, that the failed closed-loop system remains stable and its performance is as close as possible to the performance of the closed-loop system with only total faults present in the system. In cases of partial faults the second mode is practically an extension of the modified pseudo-inverse method. In cases of component faults the second mode is based on an LMI optimization problem. The approach is illustrated using a model of a real-life space robot manipulator, in which total, partial and component faults are simulated.