非平稳在有色噪声作用的参数估计与自校正控制——不平稳随机扰动作用下的鲁棒参数估计

指数遗忘的最小的二乘法(RLS)用于受到时变方差扰动的ARMAX模型的参数估计时,存在严重的参数发散问题。本文针对此类恶劣情况(Worst case),通过引入外部激励,以改善估计参数的收敛性能,从而提高估计器的鲁棒性。对引入激励的抗噪声作用进行了理论推导和仿真研究。结果表明,所用方法能提高估计器能忍受的噪信比,加速估计器的收敛。

Parameter Estimation and Self-Tuning Control Inthe Presence of Non-Stationary Colored Noise:Bobust Parameter Estimation in the Presence of Non-Stationary Disturbance

The use of Recursive Least-squire (RLS) algorithm to the parameter estimation of the Exogenous Auto-regressive Moving-average (AMARX) process model corrupted by disturbance of time - varying variance may face severe difficulty of parameter estimate diverge. This paper presents the results of an investigation on the convergency of parameter estimate of self-tuning control in.the presence of this type of worst case disturbance of jumping variance . Theoretcal results expose that the introduction of external excitation by input perturbation increases the robustness of the parameter estimator by improving the convergncy of the parameter extimates. Digital simulation results demonstrated that the introduction of external excitation increaes the noise-to - signal ratio that the parameter estimator is able to tolerate and also speed up the convergency of the parameter extimator and thus of the self-tuning controller.