A robust meta-heuristic adaptive Bi-CGSTAB algorithm to online estimation of a three DoF state-space model in the presence of disturbance and uncertainty

Most control systems require a fairly accurate model of dynamic system to design or implement a controller. When system dynamics change, the dynamic model must undergo online or offline re-estimation. The online model estimation algorithms in the time domain, especially for large models in presence of sensor noise, model uncertainty and external disturbance are almost inaccurate and unstable. In this paper, based on the dynamic model characteristics a novel online robust meta-heuristic adaptive Bi Conjugate Gradient Stabilized (Bi-CGSTAB) algorithm is proposed to estimate the model parameters and attitude simultaneously. First, the model is estimated iteratively using the output of attitude estimation from the Kalman filter algorithm, and the attitude is estimated by the output of estimated model from the least square method. The estimation method focuses on the solving algorithm of the matrix equations of the model estimation. The online robust meta-heuristic adaptive Bi-CGSTAB method uses the information of previous iteration in the current iteration to set the solving-steps toward the local optimums. This method leads to a broader and more intelligent search in the Krylov subspace of answers. The numerical results show a higher performance, robustness and more accurate model estimation than the other stated methods in the paper. 50%

Automated summary

A novel online robust meta-heuristic adaptive Bi-CGSTAB algorithm is proposed in this paper for model parameter and attitude estimation simultaneously. This method uses information from previous iterations to set solving steps towards local optimum, leading to a broader and more intelligent search in the Krylov subspace. Numerical results show higher performance and accuracy compared to other methods discussed.