Enhanced H∞ Filtering for Continuous-time State-delayed Systems

The H∞ filtering problem for continuous-time polytopic uncertain time-delay systems is investigated. Attention is focused on the design of full-order filters guaranteeing a prescribed H∞ attenuation level for the filtering error system. First, a simple alternative proof is given for an improved linear matrix inequality （LMI） representation of H∞ performance. Then, based on the performance criterion which keeps Lyapunov matrices out of the product of system dynamic matrices, a suficient condition for the existence of robust estimators is formulated in terms of LMIs, and the corresponding filter design is cast into a convex optimization problem which can be effciently handled by using standard numerical algorithms. It is shown that the proposed design strategy allows the use of parameter-dependent Lyapunov functions and hence it is less conservative than some earlier results. A numerical example is employed to demonstrate the feasibility and advantage of the proposed design.     

A bound of conservativeness in sampled-data robust stabilization and its dependence on sampling periods

This paper investigates conservativeness of the ²-induced norm in a sampled-data robust stabilization problem. Namely, it is shown that possible conservativeness of the ²-induced norm can be bounded by difference between the best achievable performance of time-invariant continuous-time controllers and that of sampled-data controllers. Moreover, this difference is proven to approach zero as a sampling period tends to zero, which means the conservativeness above loses importance when a sampling period is small. These results are important because robust stabilization based on the ²-induced norm is computationally inexpensive compared with robust stabilization based on the non-conservative robust stability index. One key lemma plays an essential role in this paper and it should be interesting independently.     

Robust H-infmity estimation for continuous-time polytopic uncertain systems

1IntroductionAs a fundamental problem in control area,linearestimation has been extensively investigated since Kalmanpresented the optimal filteringtheory[1].It is well knownthat the Kalmanfilteringtheoryis based onthe assumptionsthat the accurate mathematical models of the consideredsystems and the statistics features of the noise inputs areknown.Inthis setting,the Kalmanfilteringis also called H2filtering,which minimizes the H2norm of the operatorfromthe external input to the estimation er…     

Robust H-infinity estimation for continuous-time polytopic uncertain systems

The design of full_order robust H_infinity estimators is investigated for continuous_time polytopic uncertain systems. The main purpose is to obtain a stable and proper linear estimator such that the estimation error system remains robustly stable with a prescribed H_infinity attenuation level. Based on a recently proposed H_infinity performance criterion which exhibits a kind of decoupling between the Lyapunov matrix and the system dynamic matrices, a sufficient condition for the existence of the robust estimator is provided in terms of linear matrix inequalities. It is shown that the proposed design strategy allows the use of parameter_dependent Lyapunov functions and hence it is less conservative than earlier results. A numerical example is employed to illustrate the feasibility and advantage of the proposed design.     

Enhanced H_∞ Filtering for Continuous-time State-delayed Systems

The H∞ filtering problem for continuous-time polytopic uncertain time-delay systems is investigated. Attention is focused on the design of full-order filters guaranteeing a prescribed H∞ attenuation level for the filtering error system. First, a simple alternative proof is given for an improved linear matrix inequality (LMI) representation of H∞ performance. Then, based on the performance criterion which keeps Lyapunov matrices out of the product of system dynamic matrices, a suficient condition for the exi...     

The influence of investor's behavioral biases on the usefulness of the Dual Moving Average Crossovers

The so called Dual Moving Average Crossovers are said to be useful signals for forecasting trends of stock prices, as one of the technical analysis methods. First, we examined the usefulness of these crossovers by using historical daily closing price data and tick by tick price data of Japanese stocks. The results revealed that these crossovers were useful as confirmatory signals for forecasting market trends. Second, we tried to identify the underlying reasons for the usefulness of the crossovers. A model, which followed the Efficient Market Hypothesis, was found to fail to generate the price fluctuation where the crossovers were useful. We then developed a model that incorporated investor's suspicion about current price validity and two famous behavioral biases: conservativeness and representativeness. We identified the mechanism that those crossovers were closely related to investor's suspicion and the behavioral biases. Kotaro Miwa: He is a Ph.D. candidate at the University of Tokyo. He is also a quantitative financial analyst and fund manager at Tokio Marine Asset Managements. He received his B.A. degree from the Faculty of Engineering at the University of Tokyo in 2001. He also received M.A. degree from the Department of Systems Science at the University of Tokyo in 2003. His current research interests include behavioral finance and financial engineering. Kazuhiro Ueda, Ph.D.: He is an associate professor at the University of Tokyo. He received his B.A. degree from the Faculty of Liberal Arts and Science at the University of Tokyo in 1988. He also received M.A. and Ph.D. degrees in cognitive science from the Department of Systems Science at the University of Tokyo in 1990 and 1993. His current research interests include cognitive analysis on scientific problem solving, adaptive human-machine interface, artificial market and behavioral finance and cognitive robotics.     

On some open problems in monotonic and conservative learning

In this paper we solve some open problems in monotonic and conservative learning.