Two-Stage Estimator for Multi-micro Sensor with Generalized Bias Derived by Unknown Input

Abstract

AbstractFor linear discrete random time-varying systems with unknown inputs, an improved two-stage Kalman filter algorithm is presented to simultaneously estimate the state and generalized deviation of the sensors in the micro-electromechanical system. Based on this algorithm, a multi-sensor fusion filter estimation is implemented. First, the offset dynamic model without unknown input is derived by dimension reduction decoupling; secondly, an auxiliary full-row rank matrix is added to decouple the bias noise from the observation noise; finally, a two-stage Kalman filter is constructed with a stateless offset filter and a state filter to estimate the state and bias. Then multi-sensor fusion filtering is performed based on this algorithm. The simulation results show that the system bias, state estimation error and root mean square error of the method and its fusion algorithm are significantly reduced. The improved accuracy proves that this method is very effective.KeywordsUnknown inputTwo-stage filterMulti-sensor fusion