Optimal state filter from sequential unknown input reconstruction: Application to distributed state filtering

Abstract

SummaryThis paper generalizes Kalman filtering with an intermittent unknown input problem to be left invertible discrete-time stochastic linear systems with zero, one, or more structural delays. Contrary to the state filtering-based system inversion where the unknown input vector is reconstructed with a time delay that is equal to the structural delay of the plant, we propose an optimal state filtering by reconstructing some linear combinations of the unknown input vector with a time delay less than the structural delay. Designed under a sequential unknown input decoupling constraint, which has never been previously studied in the literature, all presented filters are very computationally efficient. The proposed state filtering is used to solve the autonomous distributed state filtering problem in large-scale networked control systems when the unknown input vector represents interactions between subsystems and when each subsystem receives intermittent information about the interaction from unreliable networks. The stochastic stability conditions of the extended intermittent unknown input Kalman filter are established when the arrival binary sequence of packet dropouts follows a random Bernoulli process.

Fault Detection of Multi Wheeled Robot Consensus Based on EKF

“…Also in [11], presents the integration of a bank of Kalman filters with an expert system for the detection and isolation of sensor faults in mobile robots. Additionally, [12] introduces a Kalman filter designed for joint state prediction and unknown input estimation within linear stochastic discrete-time systems subject to intermittent unknown inputs in measurements. Interests in FDI for nonlinear systems have grown significantly in recent years due to the fact that most of the systems, we face in practice, are nonlinear in nature such as [13] where the FDI system is based on a single model EKF filter that generates residuals as soon as the behavior of the aircraft deviates from expected, also [14] directs its focus towards fault detection in wheeled mobile robots utilizing an EKF filter.…”

Distributed robust moving horizon estimation for multisensor systems with stochastic and norm‐bounded uncertainties

“…For multisensor systems and large-scale plants, in recent years, there exists an increasing interest on distributed filtering methods because of vast applications in different real-world problems. [22][23][24][25][26][27] In distributed methods, each sensor measures state variables, computes local estimates, and then transmits them to its neighbors through a definite network topology. Therefore, the primary subject is how to design an estimator at each sensor according to the received information from the neighboring sensors.…”