Reliable state feedback and reliable observers

New procedures for the design of reliable linear, time-invariant control systems are introduced. The reliable designs are of two classes: state-feedback designs that tolerate actuator outages, and observer-based designs that tolerate sensor outages. The design methods always yield a reliable controller, provided one exists, and provided that the resulting control systems retain stability and a prescribed H_∞-norm bound despite any outages within a prespecified subset of actuators or sensors. The development of the reliable observers assumes that the sensor failures are detected and that the observer dynamics is accordingly adjusted. An alternative output-feedback design approach is also given for which failure detection is not required; however, the reliability of the resulting designs must be verified by checking an auxiliary sufficient condition, or by analysis. An example shows that this condition is not always necessary for reliability