Event‐triggered distributed Kalman filter with consensus on estimation for state‐saturated systems

Abstract

This study investigates the problem of event-triggered distributed state estimation for discrete-time, nonlinear systems with state saturation. A Kalman-like filter is developed, and consensus is first achieved with respect to the prediction estimation. The accuracy of the computed estimation is then improved via two recursive equations. The filter gains are determined in each sensor node via utilization of only an upper bound for the common error covariance, thereby resulting in a lower computational burden. Finally, the boundedness of the estimation errors is analyzed, and a comparison of the simulation results demonstrates that the proposed filtering method outperforms a recent rival method. K E Y W O R D S distributed filter, event-triggered transmission, sensor networks, state-saturated system 1 INTRODUCTION An important aspect of multisensor systems is to design distributed filters, wherein the states of a dynamic process are cooperatively estimated based on noisy measurements obtained from several sensor nodes. In distributed filtering, information is exchanged between neighboring sensors to ensure that networked filters can arrive at an agreement (consensus) as a whole on the estimated value. Recently, besides investigating different aspects of networked systems, 1,2 the problem of distributed state estimation over sensor networks has received much attention because of its wide applications, including target tracking, 3,4 communication networks, 5 and industrial operations. 6,7 In general, approaches based on consensus strategy in distributed filtering can be divided into three main categories: (1) consensus on estimation, 8,9 in which the sum of weighted differences between the local and neighboring estimations was added to the estimator of each node; (2) consensus on measurement, 10 wherein consensus is obtained on local measurements and innovation covariance. However, it involves high communication costs and does not assure convergence and (3) consensus on information, 11 especially when local information used is uniform and average, such that the stability of the algorithm is provided at all times. It is clear that the challenges posed as a result of limited communication resources and energy supply should be considered while designing estimators for commonly used battery-operated sensors. Therefore, several studies have been published to reduce unnecessary operations for data transmission in multi-sensor systems by using an event-triggered communication protocol. 12-16 For instance, in the work of Song et al, 15 a distributed extended Kalman filtering problem was suggested for discrete-time nonlinear systems with multiple fading measurements, where an event-triggered communication scheme is used in both the sensor-to-estimator and estimator-to-estimator channels. In the paper

Scalable Distributed State Estimation for a Class of State-Saturated Systems Subject to Quantization Effects

“…Remark 4: For computing the parameters of the proposed filter, the common, prediction and filter error covariances are computed in Eqs. ( 16)- (18). According to Eqs.…”

Partial‐nodes‐based state estimation for state‐saturated complex networks with random mixed couplings under dynamic event‐triggered protocol

“…[10][11][12][13] However, the state variables may be limited in a bounded set during solving such problem due to the device suffers some physical restrictions and so on. 14 As it is well recognized, fixed saturation will lead to nonlinearity characteristics, which would severely restrict the application of existing SE scheme, 15,16 which motivates the current study on the state-saturated SE problem. In Reference 17, the fusion SE strategy has been proposed for discrete CNs under state saturation.…”

Scalable recursive state estimation for nonlinear two‐dimensional complex networks with switching topology

“…The problem of state estimation has attracted considerable attention for 1-D systems with different cases over the past two decades. [9][10][11][12][13][14] In Reference 15, an event-triggering L 2 − L ∞ state estimation problem has been studied for a class of continuous stochastic neural networks subject to time-varying delays. A memory event-trigger mechanism is used to schedule the information propagation between sensor and estimator for reducing communication frequency and saving energy.…”