Compensated closed-loop Kalman filtering for nonlinear systems

In this paper, we introduce new state estimation algorithm for nonlinear systems with the measurement loss at output channels. The goal of proposed methodology is to compensate for the output state-measurement losses through an advanced scheme that relies on the nonlinear prediction (NLP) and Kalman filter. Using NLP subsystem, based on Exponential Autoregressive (EXPAR) model, a compensated observation signal is first regenerated and then it is delivered to nonlinear discrete filtering or linearized Kalman filter in the form of closed-loop. Two proposed scheme versions (EXPAR-Substitution and EXPAR-Elimination) are presented in this paper. A two-phase permanent magnet synchronous motor (PMSM) case study is exposed to loss of measurements and then the loss is reconstructed using EXPAR-based schemes to demonstrate some capable results of the proposed algorithm. Simulation results of the proposed algorithm are found to be superior compared to other algorithms in the literature. (C) 2019 Elsevier Ltd. All rights reserved.