巻号一覧
30 巻, 8 号
選択された号の論文の17件中1～17を表示しています
- |<
- <
- 1
- >
- >|
-
丁 懐東, 出澤 正徳, 松本 晋1994 年 30 巻 8 号 p. 883-891
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーA semiconductor image position sensitive device (PSD) is widely used in various measuring systems because of its simplisity of instrumentation and high speed sensing capablity of the position of a light spot as its gravity center. In order to increase the image position sensing precision, the dividing resistance should be manufactured in high accuracy. However, it is very difficult to manufacture uniformly distributed dividing resistance layer in the conventional type PSD. In order to overcome the above difficulty, a concentrated dividing resistor type PSD, in which dividing resistor is constructed separately from the photo-sensitive elements as a concentrated resistor and separately formed photo-sensitive elements are connected to it. A dividing resistor can be manufactured stably and correctly by this scheme. A comb structured PSD is proposed and trialy manufactured as a realization of the new scheme. On the contrary, the comb structure may bring the deviation in detecting characteristics which are caused by its periodic structure. Because of this reason, the interval of teeth is one of the important factors in designing the comb structured PSD. In this study, image position sensing characteristics of the comb structured PSDs which had been manufactured in trial are measured in connection with the size of the light spot. Then the influence of the light spot size to the image position sensing characteristics have been analized. Considering the above experimental results, several methods to decreasing the deviation of image position sensing characteristics caused by the comb teeth structure have been investigated and proposed.抄録全体を表示PDF形式でダウンロード (1968K) -
高橋 信, 北村 雅司, 吉川 栄和1994 年 30 巻 8 号 p. 892-901
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーThe realtime cognitive state estimator based on the set of physiological measures has been developed in order to provide valuable information on the human behavior during the interaction through the Man-Machine Interface. The aritificial neural network has been adopted to categorize the cognitive states by using the qualitative physiological data pattern as the inputs. The laboratory experiments, in which the subjects' cognitive states were intentionaly controlled by the task presented, were performed to obtain training data sets for the neural network. The developed system has been shown to be capable of estimating cognitive state with higher accuracy and realtime estimation capability has also been confirmed through the data processing experiments.抄録全体を表示PDF形式でダウンロード (3359K) -
小林 康秀, 沖田 豪, 魚崎 勝司1994 年 30 巻 8 号 p. 902-907
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーIn this paper, we consider the design of an optimal identification input for discrete-time linear dynamical systems with unknown structures. Firstly, models with distinct structures are hypothesized, following which, the concept of entropy is used to express the uncertainty of these model structures. The observations of the input-output data from the system can be used to reduce the entropy sequentially. The Kullback's divergence for the various model structures gives the upper bound of the expected decrease of the entropy.
The optimal input is the one which maximizes the Kullback's divergence at each period. This optimal input is given as the solution of a non-linear optimization problem. An approximate solution for the problem is also given which requires significantly less computation.
It is demonstrated by simulation that the optimal input can discriminate the true system structure rapidly, compared to a random input. The estimation error of the parameters and the prediction error using the optimal input signal are also superior to the random input case. The proposed method is more effective when the observation noise is relatively large.抄録全体を表示PDF形式でダウンロード (1463K) -
内田 主幹, 豊田 幸裕, 中村 秀雄, 大北 芳久, 熊丸 耕介, 村田 純一, 和田 清, 井上 勝裕, 相良 節夫1994 年 30 巻 8 号 p. 908-916
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーA method is proposed for identification of nonlinear systems. The method is primarily intended for model based failure diagnosis of large scale nonlinear plants.
A model of the plant to be identified is exploited as a simulator; the integrated squared output errors between the plant and the simulator over a short time interval is minimized by adjusting the simulator's parameters. The simulator parameters are adjusted iteratively, by small amounts at a time; in each iteration, linearization technique is applied and therefore there needs no distinguishment between linear and nonlinear plants. The initial state values are also included in the parameter vector to be adjusted; we can make the output error zero without waiting for the initial state response to decay and thus quick identification is possible. Still more important benifit is that the identifiability condition imposed on the plant input becomes weaker.
With these features, the proposed approach is called simulator-based approach to neat and quick system identification or SANQ system identification for short.
A simulation study of a feed water heating system failure diagnosis in a thermal power plant shows the validity of the method.抄録全体を表示PDF形式でダウンロード (2440K) -
松下 昭彦, 土谷 武士1994 年 30 巻 8 号 p. 917-925
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーIn this paper, we propose an on-line planning method for a desired signal making use of on-line information of a control system.
On-line planning has the advantage of off-line planning in the following points.
In the off-line planning, a servo system deals with the influence of initial state values, disturbance and parameter perturbation. In the on-line planning, however, both the planning of the desired signal and the servo system deal with the influence. That is, the influence is reduced by changing the desired signal within its permitted domain.
However, processing time for the on-line planning should be shorter than that for the off-line planning. Therefore, the planner should be simple without any trial and error. The whole system with the planner and the control system should be stable because the planner uses the state of the control system.
In this paper, we propose an on-line planning method as follows. The permitted domain of the desired signal is expressed as parameters of the desired signal. By means of this expression, the planning and the design of the control system are replaced by the design for the controlled object having limitation of the desired signal.
This method has the following merits.
(1) Many methods proposed for design of control systems can be applied because the planning is included in the design of the control system.
(2) Processing time for the planning is short because the control rule of servo system is generally simple without any trial and error.
(3) The whole system with the planner and the control system is stable.
(4) Path planning is an expanded trajectory planning problem because a permitted domain of the desired signal for the trajectory planning is expressed as the scalar parameter of the desired signal and the domain for the path planning is expressed as parameters.
Further, this method is applied to a mobile robot and its simulation results are shown.抄録全体を表示PDF形式でダウンロード (1716K) -
松野 文俊, 畑山 満則, 千田 未顕, 石邊 知明, 坂和 愛幸1994 年 30 巻 8 号 p. 926-935
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーIn this paper, modeling and vibration control of a flexible solar array paddle are considered. We first derive a partial differential equation and a set of boundary conditions of the flexible solar array paddle and ordinary differential equations of angles of rotation of motors. On the basis of a finite-dimensional modal model, an optimal controller with low-pass property and a robust H∞ controller for the flexible solar array paddle are constructed. Simulations and experiments have been carried out.抄録全体を表示PDF形式でダウンロード (1945K) -
梅本 敏孝, 青島 伸治1994 年 30 巻 8 号 p. 936-942
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーAdaptive nonlinear digital filter is proposed to eliminate sinusoidal noise from picture. Hitherto, Wiener filter optimized by original picture has been used for such purpose. The original picture, which may not be available, is not necessary for our new method. In our approach, we combine an ε-filter with an LMS adaptive algorithm. Simulations show the effectiveness of our approach.抄録全体を表示PDF形式でダウンロード (2275K) -
岡田 徳次, 野口 昌彦, 藤原 茂1994 年 30 巻 8 号 p. 943-952
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーThis paper describes calculation methods to determine an optical reflected position on a mirror ball. Optical measurement is useful to develop a smart sensor for measuring small displacement, since it has a lot of merits to be free from electromagnetic noise, temperature, in addition to its high response and much information of transmission. Our idea is to apply the measurement to a direction-of-action sensor utilizing optical reflection on a mirror ball in a spherical vessel. For this purpose, geometry of the optical reflection is analyzed so that an appropriate calculation procedure of the reflected position can be obtained.
It is shown that the reflected position is given by solving six degree linear equation. To get the solution, a method of numerical iteration is proposed. Also, three approximation methods are introduced to get the position uniquely. These are called 1) concentric tangential method, 2) two-step angular bisection method, and 3) simultaneous angular bisection method. Calculated results are compared with the most likely data by numerical iteration to show that the concentric tangential method gives the position with minimum error in the three methods.抄録全体を表示PDF形式でダウンロード (1724K) -
最上 義夫, 馬場 則夫, 反田 幸男1994 年 30 巻 8 号 p. 953-958
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーFor hierarchical structure learning automata operating in a nonstationary random environment, in this paper, a new learning algorithm is constructed by extending the relative reward strength algorithm proposed by Simha and Kurose. The learning propertiy of our algorithm is considered theoretically, and it is proved that the path probability of the optimal path can be approached 1 as much as possible by using our algorithm. In numerical simulation, the number of iterations of our algorithm is compared with that of the hierarchical structure learning algorithm proposed by Thathachar and Ramakrishnan, and it is shown that our algorithm can find the optimal path after the smaller number of iterations than that of the algorithm of Thathachar and Ramakrishnan.抄録全体を表示PDF形式でダウンロード (1035K) -
梅本 敏孝, 青島 伸治1994 年 30 巻 8 号 p. 959-965
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーIn the previous paper, we proposed an effective spectrum analysis method for transcription, in which adaptive technique was used. Step size parameter of this method is within the range of 0<μ<1/4L (L: number of weight). But, step size parameter is a convergence factor which determine adaptation rate. If large number of weight is necessary, it limits the adaptation rate. In this reason, we proposed Adaptive DFT by LMS algorithm. By the proposed method, the range of μ is 0<μ<1/4. Simulation and experimental results are illustrated.抄録全体を表示PDF形式でダウンロード (934K) -
中村 孝, 和久津 拓也, 相吉 英太郎1994 年 30 巻 8 号 p. 966-975
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーIn the asynchronous state transition of the Hopfield's type of neural networks with binary states, the state transition of neurons is trapped at one of local optima in a neighbourhood with radius of one Hamming distance, because all the transition occurs between two states at a distance of a single bit. In this paper, we present a neural network whose states transit directly through the Hamming distance of several bits and get off from such a local optimum in order to reach deterministicly the global optimum.
Concretely, the only when the states are trapped at a local optimum in the asynchronous transition mode, the mode is changed into the linked transition mode in which some of the neurons change the states cooperatively and simultaneously according to threshold rule for total inputs value concerned with the linked neurons. The simulation results for unconstrained types of 0-1 combinatorial optimization problems with a quadratic function demonstrate the fundamentals of the proposed linked state transition.抄録全体を表示PDF形式でダウンロード (1760K) -
羽深 嘉宣, 相吉 英太郎1994 年 30 巻 8 号 p. 976-983
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーA neural network is expected to have not only interpolating ability for non-training data but also noise-resisting ability which means that the outputs keep the desired signals against the perturbation from the standard input data. In order to obtain the interpolating ability, we formulated the optimization problem which requires minimization of the absolute values of synaptic weights under the satisfactory conditions that the output errors in response to the training input data are less than the permissible levels.
In this paper, the satisfaction principle is adopted to obtain the noise-resistibility also. That is, satisfactory conditions in the formulated problem require that output errors in response to arbitrary perturbation of inputs within a certain range are reduced below permissible levels. In order to solve the problem, “Relaxation Learning Method”is proposed. This learning starts by solving a relaxed problem constrained by a few satisfactory conditions corresponding to the specified input data with some noises, and the new satisfactory conditions for other kinds of input data are added to the set of the constraints iteratively. Lastly, the learning terminates when the necessary and sufficient number of satisfactory conditions to acquire the resistibility are generated.
The application to the simple pattern recognition problem demonstrates that the proposed learning method is effective for interpolating ability together with noise-resistibility.抄録全体を表示PDF形式でダウンロード (1842K) -
石川 信行, 藤井 義雄, 篠原 慶邦1994 年 30 巻 8 号 p. 984-986
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーPDF形式でダウンロード (398K) -
鈴木 達雄, 河野 通夫1994 年 30 巻 8 号 p. 987-989
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーThis paper proposes a method of closed loop eigenvalues assignment of linear periodic discrete time systems by using infrequent observations.
A necessary and sufficient condition for the closed-loop eigenvalues to be assignable arbitrarily is also given.抄録全体を表示PDF形式でダウンロード (355K) -
武藤 康彦1994 年 30 巻 8 号 p. 990-992
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーA lower left triangular interactor has been used in the multivariable control systems. But, since the triangular form is not essential for a polynomial matrix to be an interactor, we focus our attention on the more general class of interactor and its properties in this paper. It will be shown that this general class of interactors and the state equation of the inverted interactorized closed loop system are invariant under a state feedback. The relationship between these invariant properties and the maximally unobservable subspace of the feedback system will be also discussed.抄録全体を表示PDF形式でダウンロード (339K) -
愛田 一雄, 杉本 真正, 梅内 晴成1994 年 30 巻 8 号 p. 993-995
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーThe type-1 servo system with preview actions is designed to be optimal for a new performance index, which comprises the quadratic forms of the error vector and the deviation control vector, by taking into account of a trade-off between low-frequency sensitivity and guaranteed stability margin.抄録全体を表示PDF形式でダウンロード (352K) -
早川 義一1994 年 30 巻 8 号 p. 996-998
発行日: 1994/08/31
公開日: 2009/03/27
ジャーナル フリーH2-optimal sampled-data control problem is solved by deriving an equivalent H2 discrete-time control problem, where the generalized plant P(z) has a state-space representation with D21=0. This paper considers the obtained discrete-time problem and gives H2-optimal controller in state-space form.抄録全体を表示PDF形式でダウンロード (355K)
- |<
- <
- 1
- >
- >|