鲁棒递推阻尼最小二乘算法

Robust Recursive Damped Least Square Algorithm

递推最小二乘法是工程上自校正控制中常用的参数辨识方法,收敛速度快,但是存在着长期运行协方差矩阵容易爆发的问题。工业上通常采用死区等保护手段,但是会对精度产生一定的影响,并且参数调试困难。本文根据目前计算机的处理能力,提出了阻尼最小二乘法的有限时间窗口批处理递推算法,解决了传统递推最小二乘算法存在的难以克服的参数爆发现象,仿真结果验证了这个算法的有效性。

Recursive Least Square (RLS) method is common used in self-tuning control in engineering. It converges with high speed, however it may bring about eruption of the covariance matrix in a long term. Dead-zone method is usually adopted to avoid this phenomenon, but it may affect precision more or less and correlative parameters are hard to debug. Based on the processing capacity of current computers, this paper presents batch recursive damped least square method using input-output data in finite time window. It avoids parameters eruption in routine RLS .The simulation results demonstrate its effectiveness.