A fault-tolerant sensor reconciliation scheme based on LPV Unknown Input Observers

This paper presents a fault-tolerant sensor reconciliation scheme for systems equipped with a redundant number of possibly faulty “physical” sensors. The reconciliator is in charge to discover on-line, at each time instant, the possibly faulty physical sensors and exclude their measures from the generation of the “virtual” sensors, which, on the contrary, are supposed to be always healthy and suitably usable for control purposes without requiring the reconfiguration of the nominal control law. Amongst many, the solution proposed here is based on the use of a Linear Parameter Varying Unknown Input Observers (LPV-UIO) coupled with an “ad-hoc” parameter estimator used to identify on-line the current sensor reconciliation matrix. The latter is therefore used to hide the faulty measures from the pool of physical outputs in the generation of the virtual outputs. For simplicity, the sensor faults here considered are limited to variation of sensors' gain and offset values. The scheme is fully described and all of its properties investigated and proved. Finally, a simulation example is reported in details to show the effectiveness of the scheme.