Improvement of compensated closed-loop Kalman filtering using autoregressive moving average model.

• Data loss in the process of state estimation has been addressed in this article. • The existing method based on AR model has been improved by replacing ARMA model. • A simple minimum error based mechanism is proposed to decide order of LP filter. • The computational time is reduced by avoidance of inversion of matrices. • The existing schemes are compared with the proposed one to show the performance. Data loss problems severely effect the performance of state estimation in most of communication and control systems. The normal techniques adopted for compensation in the process of state estimation are Open-loop Kalman filter and compensating closed-loop Kalman filter. The compensated closed-loop Kalman filtering scheme employs three (03) strategies namely Normal Equation, Levionson-Durbin and Leroux-Gueguen algorithms, using Autoregressive (AR) model, where only previous measurements are used. In this paper, the compensated vector is proposed based on Autoregressive Moving Average or ARMA model instead of AR model. This model contains more information than Autoregressive model, i.e. measurement and input signals, which is believed to generate more efficient results. Necessary steps including the computation of linear prediction coefficients have been taken to accommodate the input signal. Computation of this extra linear prediction coefficient however, bears an observable increase in computation time. The selection of AR and ARMA models is a trade-off between computation time and improved performance, which is the ultimate consequence of extra input signal. A standard mass-spring-damper case study is simulated to provide a comprehensive comparison in light of various parameters including state estimation, error, gain elements etc. [ABSTRACT FROM AUTHOR]