This paper presents a novel approach for the identification of continuous-time systems. The proposed method is based on the projection of measured signals onto the finite-dimensional signal subspace whose basis is determined by the structure of the parameterized model of the target system. This structure-dependent basis enables us to estimate the specific parameter of the target system efficiently, and this aspect is especially useful in continuous-time system identification because parameters of such systems often correspond directly to physical parameters of the target system. Also, the proposed method can handle the closed-loop setting, which is frequently demanded in practical applications, as the specific structure of the model. The effectiveness of the method is demonstrated through a numerical example of closed-loop identification for an unstable magnetic levitation system with known physical constants.

抄録全体を表示

This paper describes overview of a model-based B2B (batch to batch) control for an industrial batch polymerization process. In order to control the reaction temperature precisely, several methods based on the rigorous process dynamics model are employed at all design stage of the B2B control, such as modeling and parameter estimation of the reaction kinetics which is one of the important part of the process dynamics model. The designed B2B control consists of the gain scheduled I-PD/II²-PD control (I-PD with double integral control), the feed-forward compensation at the batch start time, and the model adaptation utilizing the results of the last batch operation. Throughout the actual batch operations, the B2B control provides superior control performance compared with that of conventional control methods.

抄録全体を表示

It is well known that plants with time delay are hard to be controlled by using traditional method. For this, controller with delay, such as Internal Model Control (IMC), Smith-method, have been proposed for input-delay systems. However, it would be difficult to realize the delay of controller because of memory limit of micro control unit(MCU). Also, the sampling time might be large in case of the application to the plant with large time delay, because of the limitation of the memory in MCU. Hence, the trade-off exists between sampling time and maximum quantizing error, and the assignment of the quantizer affects the quantization error. In this paper, dynamic quantizers are designed for achieving small quantizing error for input-delay control systems in MCU system. Also, the attainable performance caused by assignment of the quantizer is discussed. The effectiveness of the proposed method is shown by numerical example.

抄録全体を表示

In this paper, a limiting property of full information H_∞ control is discussed. Existence of a solution for full information H_∞ control problems and limiting properties of a closed-loop system i.e. disturbance attenuation and disturbance decoupling are proposed. Moreover, relationship between limiting properties of LQ control and full information H_∞ control is mentioned. The paper also gives numerical examples to verify the limiting properties of full information H_∞ control.

抄録全体を表示

In stabilization problems of linear parabolic boundary control systems, we propose the possibility of reducing the number of the sensors (possibly the minimum) for stabilization, based on a unified algebraic approach, which remains one of the main interests in control theory.

抄録全体を表示

In this paper, a search method for a fixed-order controller that satisfies the standard H_∞ control problem is proposed. This is a descent method that gives a sequence of the stabilizing controllers so that the H_∞ performance function may be monotonically non-increasing. A frequency-dependent LMI(linear matrix inequality) with respect to the coefficient matrices of the controller is derived from the H_∞ norm constraint. This method is applicable to the frequency response of the plant.

抄録全体を表示

This paper argues how a compact humanoid robot can acquire a giant-swing motion without any robotic models by using Q-Learning method. Generally, it is widely said that Q-Learning is not appropriated for learning dynamic motions because Markov property is not necessarily guaranteed during the dynamic task. However, we tried to solve this problem by embedding the angular velocity state into state definition and averaging Q-Learning method to reduce dynamic effects, although there remain non-Markov effects in the learning results. The result shows how the robot can acquire a giant-swing motion by using Q-Learning algorithm. The successful acquired motions are analyzed in the view point of dynamics in order to realize a functionally giant-swing motion. Finally, the result shows how this method can avoid the stagnant action loop at around the bottom of the horizontal bar during the early stage of giant-swing motion.

抄録全体を表示