計測自動制御学会論文集
Online ISSN : 1883-8189
Print ISSN : 0453-4654
ISSN-L : 0453-4654
34 巻, 1 号
選択された号の論文の8件中1~8を表示しています
  • 築山 俊史
    1998 年 34 巻 1 号 p. 1-8
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    This paper describes a method for identifying planar surfaces and cylindrical objects in a scene and getting their geometrical information; the orientation and perpendicular distance to a plane and the radius and center of a cylinder. The vision system consists of a TV camera and two flash lamps. The lamps are switched on alternately, and images under each lighting are taken at two different points. Under a point light source, even on a planar surface, there is a distribution of luminance due to diffuse and specular reflections. If the two light sources are arranged at appropriate positions with respect to the camera, the positions of the peak of the luminance distribution on each surface directly give such geometrical information on the surface. Our approach is based on measuring the peak positions. This method promises to speed up the acquisition of geometrical information on an entire scene considerably, because the geometrical information on target objects can be obtained without analyzing range maps. Since the equipment setup is very simple, the proposed technique would be useful, for example, for real-world robotic applications such as navigation of indoor mobile robots. The experimental results show that the method is sufficient for such purposes.
  • 原田 博之, 西山 英治, 柏木 濶
    1998 年 34 巻 1 号 p. 9-12
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    Many classes of non-linear systems can be represented by Volterra kernels. In order to obtain the Volterra kernels, one of the authors has proposed a new method by use of M-sequences and correlation function.
    In this paper, the authors propose a new method for polynomial approximation of non-linear system using the Volterra kernels measured by the M-sequence method. The non-linear system can be approximated as a combination of a linear system and its squared portion. The parameters, which represent the relative magnitude of the squared part, is calculated by use of the Volterra kernels.
    The proposed method is applied to a non-linear chemical process. From the results of simulations, it is shown that the output signals of the non-linear system calculated by the proposed method agree well with the theoretical outputs.
  • 津村 幸治, 森田 洋, 齋藤 義夫
    1998 年 34 巻 1 号 p. 13-19
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    Continuous time deadbeat control is eliminating controlled signal of a continuous time system in a finite time. In this paper we show an application of continuous time deadbeat control: deadbeat control of continuous time H control system. To design a deadbeat control system, we use a free parameter function of a controller in a standard H control system. The free parameter function should satisfy simultaneously some interpolation conditions for deadbeat control and a norm condition for H control. We parametrize the functions which satisfy the interpolation conditions and find a function satisfying the norm condition with linear matrix inequality. We also give some numerical examples and show the effectiveness of the proposed design.
  • 二輪車の安定化制御
    村田 剛, 川谷 亮治
    1998 年 34 巻 1 号 p. 20-26
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    Loop Shaping Design Procedure (LSDP) is one of the most promising H design procedure. However, LSDP has a weak point such that the so-called central controller has higher order. In this paper, we show that for a single input system which has a structure of mechanical system, there exists a lower order controller by selecting free parameter appropriately and the controller has a structure of functional observer. And we apply our design procedure to the stabilization problem of a two-wheeled vehicle which is a typical unstable mechanical system.
  • 田中 秀幸, 杉江 俊治
    1998 年 34 巻 1 号 p. 27-33
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    This paper is concerned with simultaneous design of structure parameters (passive components) and controller (active components) in order to achieve optimal system performance. First, a fairly general framework for the simultaneous design problem is proposed based on LFT (Linear Fractional Transformation). Second, it is shown that the problem can be reduced to BMI (Bilinear Matrix Inequality) problem, using appropriate descriptor form representation. Furthermore, we give a computation method to solve BMI problems which is suitable for the simultaneous design problem, and evaluate its effectiveness via numerical example.
  • 山田 克彦, 吉河 章二
    1998 年 34 巻 1 号 p. 34-40
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    The feedback control of attitude changes for a spacecraft with two reaction wheels is considered in this paper. The method used here is to utilize the phenomenon that an attitude change occurs around an axis perpendicularly intersecting the wheel rotation axes when the two reaction wheels rotate simultaneously. A periodic feedback law is proposed for the three-axis control of spacecraft attitude. The validity of the law is verified by numerical simulations.
  • 古賀 勝, 平澤 宏太郎, 大林 正直
    1998 年 34 巻 1 号 p. 41-47
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    In this paper, evaluation between Likelihood Search Method (L.S.M.) and Back Propagation Method (B.P.M.) in Neural Networks (N.N.) learning is studied. The L.S.M. is an optimization method which can search for a global optimum systematically and effectively in a single framework, which is not a combination of different methods. The L.S.M. can realize the intensification and diversification of the search based on an idea that the searching for variables is intensified where a likelihood of finding good solutions is high, on the other hand, the searching for variables is diversified where the likelihood is low. The L.S.M. is a sort of Random Search Method (R.S.M.) but utilizes gradient information, and the likelihood of finding good solutions is defined by a norm of gradient. In simulations, the learning ability is evaluated between the L.S.M. and B.P.M. in N.N. learning. Simulations are carried out to realize nonlinear functions by using L.S.M., B.P.M. and moment B.P.M. in a layered N.N.. The simulation results show that the L.S.M. is superior to the B.P.M. because of the ability of intensification and diversification of the search.
  • 佐藤 祐司, 石井 弘允
    1998 年 34 巻 1 号 p. 48-54
    発行日: 1998/01/31
    公開日: 2009/03/27
    ジャーナル フリー
    The avoidance of ship collisions becomes an important issue for maritime traffic safety with the increase of traffic density in the inland sea and coastal region.
    We have evaluated the risk of ship collisions during meeting through the analyses of radar and IR image information.
    This paper describes identification of the category of a target vessel from IR image. Feature parameters describing the size and shape of the vessel are employed for identification of the vessel category on the basis of the three-dimensional graphic models of representative vessels. These parameters evaluated for each category are learned by a multilayered neural network. The outline of a vessel is extracted from IR image and the feature parameters are inputted to the neural network to examine identification of the vessel category. The result reveals that the identification can be done satisfactorily and the present system can effectively be applied to the avoidance of ship collisions.
feedback
Top