基于目标函数的递推子空间辨识算法

针对实际工程中要求对系统参数进行在线估计的问题, 提出一种递推子空间辨识的新方法. 通过引入辅助变量关系将递推子空间辨识问题转化为目标函数的迭代最小化问题. 采用递推最小二乘算法在线估计参数并由传播方法得到更新的广义能观性矩阵, 进而求得子空间辨识模型系统参数. 该算法简单有效且对初值具有鲁棒性. 最后, 通过仿真实例验证算法的有效性.

Recursive subspace identification algorithm based on objective function

We propose a new recursive subspace identification algorithm for online estimating the system parameters in practical engineering. By introducing the instrumental variable, we convert this problem into a recursive minimization problem of an objective function. The recursive least square algorithm is employed to estimate the parameters, then the propagator method is developed to update the extended observability matrix; and the system matrices are finally computed. The proposed method is simple and effective and is improved to be robust to the uncertainty in initial values. The efficiency of this method is illustrated with a simulation example.