This paper is concerned with quantized feedback control in the case where logarithmic-type dynamic quantizers are adopted instead of conventional static (memoryless) ones. First, when the plant and the state feedback controller are given, the admissible coarsest quantization density which guarantees quadratic stability of the closed loop system is given in a closed form, which does not depend on the choice of controller in contrast to the static quantizer case. Second, when the plant, the state feedback controller and the coarseness of the quantization density are given, we provide a design method of the dynamic quantizers via convex optimization. Third, these results are extended to the case of output feedback control systems. Finally, some numerical examples are given to demonstrate the effectiveness of the proposed method.

抄録全体を表示

In the present work, a new practical method for direct tuning of PID controllers using operation data under feedback control is proposed. Conventional direct tuning methods have the following problems: need for iterative experiments and difficulty in properly determining a reference model without information on a process. The objective of this research is to propose Extended Fictitious Reference Iterative Tuning (E-FRIT) for solving these problems through 1) the objective function is modifieded to include the penalty for changes of the input variable, 2) the parameter in the reference model is optimized together with PID control parameters, and 3) particle swarm optimization (PSO) is used to make the first two extensions with facility. The usefulness of the proposed E-FRIT is demonstrated through the case studies of unstable chemical reaction processes.

抄録全体を表示

Biometrics is classified into verification and identification. Many researchers on the keystroke dynamics have treated the verification of a fixed short password which is used for the user login. In this research, we pay attention to the identification and investigate several characteristics of the keystroke dynamics in Japanese free text typing. We developed Web-based typing software in order to collect the keystroke data on the Local Area Network and performed experiments on a total of 112 subjects, from which three groups of typing level, the beginner's level and above, the normal level and above and the middle level and above were constructed. Based on the identification methods by the weighted Euclid distance and the neural network for the extracted feature indexes in Japanese texts, we evaluated identification performances for the three groups. As a result, high accuracy of personal identification was confirmed in both methods, in proportion to the typing level of the group.

抄録全体を表示

Supervisory control is a general framework of logical control of discrete event systems. A supervisor assigns a set of control-disabled controllable events based on observed events so that the controlled discrete event system generates specified languages.
In conventional supervisory control, it is assumed that observed events are determined by internal events deterministically. But, this assumption does not hold in a discrete event system with sensor errors and a mobile system, where each observed event depends on not only an internal event but also a state just before the occurrence of the internal event. In this paper, we model such a discrete event system by a Mealy automaton with a nondeterministic output function. We introduce two kinds of supervisors: one assigns each control action based on a permissive policy and the other based on an anti-permissive one. We show necessary and sufficient conditions for the existence of each supervisor. Moreover, we discuss the relationship between the supervisors in the case that the output function is determinisitic.

抄録全体を表示

This paper considers linear time-invariant continuous-time systems with control input saturation nonlinearities, and proposes two regional stability synthesis methods of output feedback controllers such as an ellipsoid defined by a level set of a quadratic Lyapunov function to be a domain of attraction for the systems based on the generalized sector approach. One of them is an integrated design of full-order dynamical output feedback and anti-windup controllers with the same plant order, while the other is an integrated design of reduced-order dynamical output feedback and antiwindup controllers to be less than the number of available plant states from the plant order. The two methods assume the output of the nonlinearities to be available for the control. In this case, this paper indicates that the synthesis problems using the two methods can be recast as linear matrix inequality (LMI) optimization problems respectively. Furthermore, it is proved that two subsets of achievable domains of attraction using the two controllers are exactly the same. Thus this paper concludes that the reduced-order controller does not decrease the size of the achievable domain of attraction, within our framework, when compared with that resulting from the full-order controller.

抄録全体を表示