极小极大拟合准则下的在线参数估计方法

On-line Parameter Estimation with the Minimax Fitting Criterion

提出一种采用极小极大拟合准则的实用的在线参数估计方法。为此采取了两条有效措施。第一，将在线辨识中最常用的遗忘因子引入极小极大拟合准则；第二，通过研究有关的极小极大优化问题的对偶问题设计在线递推算法。前者使推广后的极小极大拟合准则具有追踪时变参数的能力，后者则为构造实用的在线递推算法提供了可行的途径。大量仿真研究表明，提出的在线参数估计方法具有较好的性能。

A practical on-line minimax parameter estimation algorithm is proposed.Two effectivemeasures are taken for such parameter estimation,namely,i) introducing into the minimaxfitting criterion the forgetting factor most frequently used in on-line identification,and ii)designing an on-line recursive algorithm by studying the dual problem of the minimax opti-mization problem,The first measure makes the extended minimax fitting criterion capable oftracking time-varying parameters while the second provides a feasible way of constructing apractical on-line recursive algorithm,A large number of simulation investigations show thatthe performance of the new algorithm is fairly good.