不需要持续激励的辨识算法及其全局收敛性

An Identification Algorithm and Its Global Convergence without Persistent Excitation Requirement

本文提出了一种带有遗忘因子和死区的数据归一化递推最小二乘辨识算法,证明了这种算法不需要持续激励具有全局收敛性,且参数收敛到真值。最后通过例子给出了仿真结果。结果表明,本文算法不需要持续激励是可行的。

A data normalization least-square identification algorithm with the forgetting factor and deadzone is proposed. With this algorithm, the estimated parameter will convergn to its true value without persistent excitation requirement, and with the deadzone, the 'bursting' phenomenon in algorithm can be avoided and the approach of identification variance F (t) to zero will be restricted. This algorithm can be used to identify a slowly time-varying system. Finally the global convergence of the algorithm is proved. Several simulation results are given.