Towards Robust Real-time System

UM > GRADUATE SCHOOL

Status	已發表Published
	Towards Robust Real-time System
	Yuen, K. V.; Mu, H.Q.; Kuok, S.C.
	2014-03-01
Source Publication	In Search of Mindset for Change for Control, Automation and Instrumentation New Technologies
Abstract	Real-time system identification has received tremendous attention in various science and engineering disciplines over the last decades. However, in practice, the quality of the identification results can be affected by possible outliers in the measurement. Furthermore, the popular Kalman filter requires a prescribed set of noise parameters and inappropriate choice may lead to unstable identification results. In this paper, a noise-parameters-updated outlier-resistant extended Kalman filter (NPU-OR-EKF) is presented for robust system identification using dynamic response data. The noise parameters in the extended Kalman filter will be updated in a real-time manner and this resolves the possible stability problem due to inappropriate assignment. On the other hand, the probability of outlier for each data point is proposed to determine its outlierness. The data points with probability of outlier over 0.5 will be regarded as outliers and they will be discarded for identification purpose. In contrast to other existing outlier detection criteria, the presented approach does not require subjective ad hoc choice of the user. Finally, the proposed NPU-OR-EKF algorithm will be applied to estimate the parameters of an oscillator with degrading stiffness using the outlier contaminant measurement.
Keyword	Kalman filter outlier detection real-time algorithm system identification
Language	英語English
The Source to Article	PB_Publication
PUB ID	27871
Document Type	Conference paper
Collection	GRADUATE SCHOOL
Recommended Citation GB/T 7714	Yuen, K. V.,Mu, H.Q.,Kuok, S.C.. Towards Robust Real-time System[C], 2014.
APA	Yuen, K. V.., Mu, H.Q.., & Kuok, S.C. (2014). Towards Robust Real-time System. In Search of Mindset for Change for Control, Automation and Instrumentation New Technologies.

Files in This Item:
There are no files associated with this item.

If you have any objections to this item, please fill out the form below and the administrator will contact you as soon as possible.
Content:
Email：	*
Affiliation No.
Verification Code:	Refresh

Any comments and suggestions are welcomed.
Title:	*
Content:
Email：	*
Verification Code:	Refresh