报告摘要: The least-squares (LS) algorithm has been widely used in system identification. However for autoregressive moving average (ARMA) model, the LS algorithm is less effective. In fact the LS estimate is biased in presence of input-output noises. We are thus motivated to propose a graph space approach in order to mitigating the bias issue. This presentation will show that system identification in presence of input-output noises involves the error-in-variable (EIV) model with a special structure. Under the additive white Gauss noise (AWGN), our proposed algorithm based on the graph space approach yields an asymptotically maximum likelihood estimate (MLE) that converges strongly to the true ARMA model. An example is worked out to illustrate our proposed graph space approach which demonstrates the convergence of the estimation variance to the corresponding Cramer-Rao lower bound (CRLB).

 

报告人简历：Professor Guoxiang Gu has performed research in system identification, robust control, and digital signal processing. He has published over 60 archive journal papers, co-authored a book, and numerous book chapters and conference publications. His research initially focused on identification and control of uncertain systems and their industrial applications. Later on he shifted his work to bifurcation control with applications to control of axial flow compressors in aeroengines. At present he focuses on estimation and control for feedback systems over the networks, including collaborative estimation issues encountered in sensor networks, feedback stabilization, and system identification with finite digits or finite channel capacity. His research has been supported by AFOSR, ARO, LEQSF, and NSF. He received Research Initiation Award from National Science Foundation (1991-1993), LSU Research Council Award (1994), and is F.Hugh Coughlin/CLECO Distinguished Professor of Electrical Engineering at LSU. He has supervised 8 Ph.D. students and 19 M.S. students since joining LSU. He is currently an Associate Editor for SIAM Journal on Control and Optimization and Automatica, and was an Associate Editor for IEEE Transactions on Automatic Control from 1998 to 2000. His current research interests include networked feedback control systems with focusing on estimation and control in uncertain environments.

 

（邀请人： 周彤，清华大学自动化系）