Decomposition of the Retrospective Performance Variable in Adaptive Input Estimation
Decomposition of the Retrospective Performance Variable in Adaptive Input Estimation
复制标题
DOI：
10.23919/acc53348.2022.9867833
复制
发表时间：
2022-06
期刊：
影响因子：
--
通讯作者：
Sneha Sanjeevini;D. Bernstein
中科院分区：
文献类型：
--
作者：
Sneha Sanjeevini;D. Bernstein
回溯成本输入估计（Retrospective Cost Input Estimation，RCIE）是一种自适应输入估计技术，它基于使用递归最小二乘法最小化回溯性能变量。在本文中，为了更好地理解RCIE的基本机制和性能，提出了一种将追溯性能变量分解为性能项和模型匹配项之和的方法。由于这种分解涉及时变的输入-输出模型，从LTV输入-输出模型的LTV状态空间实现的建设，以及从LTV状态空间模型的LTV输入-输出模型的建设。分解的分析显示RCIE如何避免收敛到不稳定或性能差的估计量。一个数值例子被用来说明推导的结果和意见。
Retrospective cost input estimation (RCIE) is an adaptive input estimation technique that is based on the minimization of a retrospective performance variable using recursive least squares. In this paper, in order to obtain a better understanding of the underlying mechanism and the performance of RCIE, a decomposition of the retrospective performance variable into the sum of a performance term and a model-matching term is presented. Since this decomposition involves time-varying input-output models, the construction of LTV state space realizations from LTV input-output models as well as the construction of LTV input-output models from LTV state space models are presented. Analysis of the decomposition shows how RCIE avoids convergence to an estimator that is destabilizing or has poor performance. A numerical example is used to illustrate the derived results and observations.