The proposed method can achieve asymptotical convergence along a learning repetition horizon.

• 随着迭代学习次数的增加，系统跟踪误差渐近收敛于零。