Fault Detection by Desynchronized Kalman Filtering, Introduction to Robust Estimation

This paper deals with the state estimation of dynamic systems. A recursive linear MMSE estimator is presented as an alternative to Kalman filtering . This estimator has the ability to cope with asynchronous measurements, and to process the data by sets of undefined sizes. It is particularly suitable for fault detection, because the decisions can be based on more data. This is an open door to robust estimation. A mixed estimator robust to various failure scenarios is then derived by using the Bayesian approach. This mixed estimator is originally thought for applications requiring high integrity estimations. It is next tested on a rail navigation problem.