An online intelligent robust adaptive LSQR estimation method for LTI state space model

Regarding the low accuracy and instability of common online methods for estimating dynamic models in the time domain, in the presence of uncertainty in system dynamics, sensor noise and environmental disturbances, this area is still open for further research. In this paper, a new estimation method is proposed based on a new online robust meta-heuristic adaptive LSQR (ORALSQR) for simultaneous estimation of a multi input/output linear dynamic model and system state variables. This new adaptive LSQR algorithm is used to solve the output matrix equations of the least squares error problem. The presented algorithm, based on its iterative nature, searches the answer subspace by using a new meta-heuristic logic. In addition, the algorithm solving steps and the search domain size in each iteration are intelligently determined by the method. In an identification maneuver, this method estimates the state variables using an estimated model in the Kalman filter, then estimates the model online for the next iteration using the state variables. In addition the stability proof of this method is presented. Numerical results show more accuracy and robustness of this method compared to the other methods mentioned in this paper which contain LS and RLS based estimation methods.

Automated summary

In this paper, a new online robust meta-heuristic adaptive LSQR (ORALSQR) estimation method is proposed for simultaneous estimation of a multi input/output linear dynamic model and system state variables. Numerical results show that this method outperforms the LS and RLS based estimation methods mentioned in this paper in terms of accuracy and robustness.