Improved estimation performance using known linear constraints
Improved estimation performance using known linear constraints
复制标题
DOI：
10.1016/j.automatica.2004.03.001
复制
发表时间：
2004-08-01
期刊：
影响因子：
6.4
通讯作者：
Söderström, T
中科院分区：
文献类型：
--
作者：
Mahata, K;Söderström, T
在与线性动态系统有关的参数辨识问题中，以线性或非线性约束的形式可获得的附加信息往往是未被开发的。我们在这项工作中的目标是探索线性约束的知识，以实现显著提高参数估计的精度。在这里讨论的这类问题中，未知的边界条件看起来像是扰民参数。在实际应用中，将这些讨厌的参数从损失函数中剔除，得到感兴趣参数的可变投影优化问题。在这项工作中，我们解决了一个约束优化问题，其中附加的线性约束是以部分已知边界条件的形式施加的。在这个过程中，我们展示了如何通过考虑约束来提高估计的精度。理论方法也被成功地应用于数值模拟和真实世界的实验中。(C)2004爱思唯尔有限公司。保留所有权利。
The additional information available in the form of linear or nonlinear constraints are often remain unexplored in the parameter identification problems related to linear dynamic systems. Our goal in this work is to explore the knowledge of the linear constraints to achieve significant improvement in the accuracy of the parameter estimates. In the class of problems being addressed here, the unknown boundary conditions appear as nuisance parameters. In practice, these nuisance parameters are eliminated from the loss function to get a variable projection optimization problem in the parameters of interest. In this work, we solve a constrained optimization problem instead, where the additional linear constraints are imposed in the form of partially known boundary conditions. In the process, we show how the accuracy of the estimates is improved by taking the constraints into account. The theoretical methodology is successfully applied also to numerical simulations as well as in real-world experiments. (C) 2004 Elsevier Ltd. All rights reserved.