Zeung nam Bien

On bounds of the Riccati and Lyapunov matrix equations

Various bounds for the traces of the solutions of the algebraic Riccati and Lyapunov matrix equations are established in the continuous and discrete domain, respectively. The presented results give lower bounds for the sum of the eigenvalue of the matrix solutions.

Proportional minus delay controller

A new type of controller, which utilizes the time-delay effect, is proposed. It is shown that the conventional P-controller equipped with an appropriate time-delay performs an averaged derivative action and thus can replace the PD-controller, showing quick responses to input changes but being insensitive to high-frequency noise.

An adaptive control scheme for robot manipulators

An adaptive control scheme is developed for a robot manipulator to track a desired trajectory as closely as possible in spite of a wide range of manipulator motions and parameter uncertainties of links and payload. The presented control scheme has two components: a nominal control and a variational control. The nominal control, generated from direct calculation of the manipulator dynamics along a desired trajectory, drives the manipulator to a neighbourhood of the trajectory. Then a new adaptive regulation scheme is devised based on the Lyapunov direct method, which generates the variational control that regulates the perturbation in the vicinity of the desired trajectory.

Intervalized iterative learning control for monotonic convergence in the sense of sup-norm

In this paper, the monotonic convergence in the sense of sup-norm is studied. First, it is pointed out that, when typical iterative learning control (ILC) algorithms are applied to LTI systems, some huge overshoot in the sense of sup-norm may be observed even though monotonic convergence in the sense of λ-norm is guaranteed. Then, new types of ILC algorithms are proposed for continuous-time systems and discrete-time systems to resolve such an undesirable phenomenon by adopting an intervalized learning scheme.

A root-locus technique for linear systems with delay

A new method of plotting the root-loci is developed for the linear control system with delay in control or in state. In case of the system with delay in control, the root-locus plot starts from neighborhoods of the open-loop zeros instead of the open-loop poles and thus the effect of time-delay is easily handled. In case of the system with delay in state, the open-loop poles are firstly found by applying the root-locus method for the system with delay in control and then the desired root-loci are found by starting the root-loci plot from the open-loop poles.

Application of neural networks to multiple alarm processing and diagnosis in nuclear power plants

A feasibility study of multiple alarm processing and diagnosis using neural networks is presented. The backpropagation network (BPN) algorithm is applied to the training of multiple alarm patterns for the identification of faults in a reactor coolant pump (RCP) system. The general mapping capability of the neural network makes it possible to identify a fault easily. A number of case studies are performed, with emphasis on the applicability of the neural network to the pattern recognition of multiple alarms. Based on the case studies, the neural network can identify the cause of multiple alarms well, although untrained, incomplete/sensor-failed or time-varying alarm symptoms are given. Also, multiple faults are easily identified with a given alarm pattern. >

A note on the stability of large scale systems with delays

A stability test is proposed for large scale systems with delays by employing both the aggregation technique based on a Lyapunov function and the strictly quasi-diagonal dominance.

A Nonsupervised Learning Framework of Human Behavior Patterns Based on Sequential Actions

In designing autonomous service systems such as assistive robots for the aged and the disabled, discovery and prediction of human actions are important and often crucial. Patterns of human behavior, however, involve ambiguity, uncertainty, complexity, and inconsistency caused by physical, logical, and emotional factors, and thus their modeling and recognition are known to be difficult. In this paper, a nonsupervised learning framework of human behavior patterns is suggested in consideration of human behavioral characteristics. Our approach consists of two steps. In the first step, a meaningful structure of data is discovered by using Agglomerative Iterative Bayesian Fuzzy Clustering (AIBFC) with a newly proposed cluster validity index. In the second step, the sequence of actions is learned on the basis of the structure discovered in the first step and by utilizing the proposed Fuzzy-state Q--learning (FSQL) process. These two learning steps are incorporated in an amalgamated framework, AIBFC-FSQL, which is capable of learning human behavior patterns in a nonsupervised manner and predicting subsequent human actions. Through a number of simulations with typical benchmark data sets, we show that the proposed learning method outperforms several well-known methods. We further conduct experiments with two challenging real-world databases to demonstrate its usefulness from a practical perspective.

Hierarchical optimal control of urban traffic networks

This paper deals with the problem of optimally controlling traffic flows in urban transportation traffic networks. For this, a nonlinear discrete-time model of urban traffic network is first suggested in order to handle the phenomenon of traffic flows such as oversaturatedness and/or undersaturatedness. Then an optimal control problem is formulated and a hierarchical optimization technique is applied, which is based upon a prediction-type two-level method of Hirvonen and Hakkala.

Feature Set Extraction Algorithm based on Soft Computing Techniques and Its Application to EMG Pattern Classification

Recognizing bio-signals, such as EMG, EEG, EOG and ECG, is a promising theme of study since it provides with a convenient means for human-machine interaction. Various approaches of determining features of bio-signals were known for discerning predefined motions/intentions of human, but most of them are applicable mostly only to a single subject, due to inherent characteristics of bio-signals. Lately, several new types of pattern classifier with known features have been proposed to cope with the problem of subject-dependency, but their error rates are still conspicuous when accommodating multiple subjects. Based on the soft computing techniques, this paper presents a comparative experimental study to minimize the subject-dependency. It is shown that the induced feature vector set obtained by the proposed algorithm has less subject-dependency than other existing methods.