-
Hisashi TSURUOKA, Kohyu FUKUNISHI
1992 Volume 28 Issue 7 Pages
777-782
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
The application of neural network is considered in solving the inverse problem at high speed in producing maps of neuron current densities, resulting from brain activity, with magnetic field data measured by superconducting quantum interference device (SQUID) sensors. Neuron activity in the brain is modeled as current dipoles. The dipoles are located in a 2-dimensional reconstruction plane which is separate but parallel with the sampling plane. Each coordinate component of a vector dipole is assigned to a neuron.
If zero is selected as the initial state of each neuron, and if the distance (=
Dd) between the reconstruction plane and the sampling plane is short, the location of dipoles is obtained correctly by a Hopfield type neural network after a small number of iterations. For the assurance of the spontaneous energy minimization process, coefficients of energy function are selected to cancel out the diagonal elements of connectivity matrix.
Boltzmann machine makes it possible to obtain the correct location with no relation to Dd and the initial state of each neuron under a priori information about the true number of dipoles.
View full abstract
-
Kazuyuki NAGATA, Tsukasa OGASAWARA, Kunikatsu TAKASE
1992 Volume 28 Issue 7 Pages
783-789
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
It is necessary to estimate a grasping pose when a robot changes its grasping state.
This paper proposes a new technique for estimating 3-D position and orientation of a grasped object with respect to the hand coordinate system using contact information. It is assumed that the grasped objects are polyhedron. In this paper, two techniques to estimate the grasping pose are described. One uses force data as contact information, and the other uses joint angles of a manipulator.
The estimation process is as follows. First, the system contacts a vertex of the grasped object with external environment. Then, the contact point with respect to the hand coordinate system is detected by using contact force data or joint angles of the manipulator. This contact point with respect to the hand coordinate system is transformed to the point with respect to the grasped object coordinate system from the internal model. The above-mentioned process is repeated several times changing the contact vertex of the grasped object. Relative rotation and translation between hand coordinate frame and grasped object coordinate frame are computed from these data using Newton method.
In the case of a parallel-jaw gripper, consideration of geometrical constraints of polyhedral model reduces the number of transformation parameters from hand coordinate frame to grasped object coordinate frame and the calculation time. Experimental results of estimating the grasping pose with the DD robot ETA-3 is also described.
View full abstract
-
Multi-Input Discrete-Time Optimal Regulator with the Fixed Terminal State Condition
Eiji KONDO, Takanori KIYOTA, Teruo SUNAGA
1992 Volume 28 Issue 7 Pages
790-795
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This paper discusses the optimal regulator problem for the multi-input linear time-invariant discrete-time system with the zero terminal state. By using the general form of the inputs which satisfy the zero terminal state condition, the system is transformed into the new time-variant system with the time-variant structure of the new input. The performance index is transformed into the sum of the quadratic forms of the new state vectors and the new input vectors with the cross terms, and the new terminal state is free. The optimal inputs are given by the variable gain state feedback in the first phase, by linear combinations of the constant gain state feedback and the variable gain state feedback in the second phase, and by the constant gain state feedback in the third phase. The variable gains are obtained by using the nonstationary Riccati equation. All gains are independent of the initial state of the system.
View full abstract
-
Naohisa OTSUKA, Hiroshi INABA
1992 Volume 28 Issue 7 Pages
796-800
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This paper studies the problem of simultaneously achieving block triangular decoupling, disturbance rejection and pole assignment by state feedback for linear multivariable systems in the so-called geometric approach, and presents necessary and sufficient conditions for this simultaneous problem to be solvable.
View full abstract
-
Tsutomu MITA, Kazunobu KURIYAMA, Kang-Zhi LIU
1992 Volume 28 Issue 7 Pages
801-809
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
Although there are many studies on the
H∞ control, little is known about structures of
H∞ state feedback controllers. In this paper we first correct the parametrization of
H∞ controllers of the FI problem derived by DGKF since they has not parametrized a null space of
T3. At the same time we also show an elementary proof of the necessary condition of the FI problem which only use the Nehari Theorem rather than the Hankel-Toeplitz operator. Using the result, we next parametrize the
H∞ state feedback controllers and explain how dynamical free parameters implied in the
H∞ controller are related with constant feedback gains which are different from the central solution
F∞. Finally we derive some property of the Riccati equation associated with the central solution.
View full abstract
-
Toshiharu SUGIE, Masahiko MURAI
1992 Volume 28 Issue 7 Pages
810-816
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This paper presents a new approach to analysis and design of multivariable sampled-data control systems. First, exploiting the modified pulse transfer function representation, we show that the factorization approach is useful to characterize all the admissible intersample output behavior in sampled-data control systems. Second, we give a solution to the
l2/L
2-induced norm optimization problem, where continuous time behavior of sampled-data systems is taken into account. Then, we apply these results to the model-matching problem to show the validity of our method by simulation.
View full abstract
-
Hideki KOKAME, Seiichi YOSHIDA, Takehiro MORI
1992 Volume 28 Issue 7 Pages
817-824
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
For systems with uncertain parameters, stability robustness is often defined as the minimum of destabilizing parameter perturbation vectors, in the sense of a predescribed norm. The stability robustness thus defined can be visualized as the radius of the largest stability hypersphere in parameter space.
The present paper considers the stability robustness, assuming a characteristic polynomial whose coefficients are an affine function of real plant parameters. Explicit formulas are established for discretetime systems. They use two polynomials in the form of weighted sum of Chebyshev polynomials, determined from the coefficient data of the affine functions. Every hyperplane which consists of the boundary of stability region in the parameter space is found from the polynomials. Furthermore the explicit formulas would provide useful means to analyse continuity and discontinuity of the stability robustness with respect to the coefficient data.
View full abstract
-
Ichiro UEDA, Katsuya KONDO, Toshiya OOI
1992 Volume 28 Issue 7 Pages
825-833
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
Requirements for the high accuracy in thickness of cold rolled strip are becoming ever more stringent and automatic gauge control systems (AGC) with high performance are also required for reversing cold mills. Moreover, reversing cold mills are turning to produce thin strip under high rolling speed in recent years. Therefore, it is necessary to design AGC systems also in terms of the dynamics in strip thickness change.
In this paper, the dynamics in strip tension response with conventional reel motor current control were examined by computer simulation. The simulation result indicates that the strip tension response to the reel motor current change is delayed especially under the conditions of high rolling speed and thin strip thickness. Therefore, the improvement of strip tension response based on the measurement of the actual strip tension is indispensable for high performance AGC systems. However, it is not practical to directly measure the actual strip tension because the maintenance of measurement accuracy is difficult under the rolling environment.
In order to solve the above problem, a new technique for estimating the actual strip tension using observer method was investigated. Based on the estimation of strip tension, a new type of strip tension control system (Active Tension Control) has been developed. In the Active Tension Control, the reel motor current reference from AGC is corrected by the accelerating or decelerating current so that the estimated value of strip tension corresponds to the strip tension reference. This system has been incorporated into the AGC system for No. 2 reversing cold mill at Wakayama Steel Works and operates effectively to provide high accuracy in thickness.
View full abstract
-
Steering and Suspension Control
Yasutaka HAYASHI, Shun'ichi DOI, Katsuhiko HATTORI, Hideo ARAKAWA
1992 Volume 28 Issue 7 Pages
834-843
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
A feasibility study of an active controlled vehicle is carried out so as to improve performances of the vehicle safety and comfortability remarkably. The purposes of this study are to develop a basic technology for realizing an actively controlled vehicle and to predict its effect on these performances.
An experimental vehicle is constructed with a front active steering system and independent hydropneumatic suspension systems which are integrated control elements and hydraulic power sources for individual wheel. A design procedure utilizing a modern control theory and a sophisticated element technique using each actuator and multi-microprocessors are also applied to the vehicle. The active steering control strategy is based on the optimal control theory including a driver-vehicle closed control system. It is possible to improve the vehicle controllability and stability at high speeds by a slight front steering angle control. On the other hand, the characteristic of the suspension system is functional integrations in each suspension unit. These systems are expected reductions of generated noise from oil pipes and spending energy to compress the oil. The active suspension control system uses combined methods of feedforward and feed back controls corresponding to various running conditions.
Experiments by real running tests show that every performance like controllability, stability, roll motion and riding comfort is improved. Comparing with controlled and noncontrolled vehicle, as well as the reduction of generated rolling angle below 50% due to steering, especially in the lane-change maneuver tests, settling time of the lateral motion of the vehicle is improved to 25% reduction and integral value of squared steering angle which show the amount of driver's steering operation is improved to 35% reduction according to simultaneously controls of both steering and suspension. The easiness of steering in driver-vehicle closed loop is derived from these results.
View full abstract
-
Independent Rear Wheel Torque Control
Norio IWAMA, Yukio INAGUMA, Katsuhiro ASANO, Yasushi AMANO, Toshinobu ...
1992 Volume 28 Issue 7 Pages
844-853
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
The purpose of this study is to research the feasibility of an integrated autombile control system, which has various control functions realized by conventional 4WS, ABS, etc. For this purpose we developed a new drive control system in which front wheels are driven by an engine, while rear wheel are driven by two electric motors. Yaw rate control yields additional moment by drive force differrence between two rear wheels. Slip control regulates the slip ratio of each rear wheel. Motor control executes high speed torque control. These controls are performed by a sophisticated multi-microprocessor system. The following results have been derived; (1) The robustness against disturbances caused by external forces was improved. (2) The acceleration and braking capabilities on low friction roads were improved. (3)The cornering stability on low friction roads was augumented by distributing the driving forces appropriately onto three drive axles.
In conclusion, this study has revealed that various automobile control functions can be integrated in a single mechanism by means of independent wheel drive system.
View full abstract
-
Toshiaki TABUCHI, Nanayo FURUMOTO, Hiroyuki FURUMOTO, Toshiro OKAHISA
1992 Volume 28 Issue 7 Pages
854-863
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This paper deals with the improvement of convergence rate or estimation accuracy of the estimates in ARMA parameter estimation by Recursive Pseudo Linear Regression (RPLR) method. For the improvement of the convergence rate, Fisher Information Matrix (FIM) is used as a measure of the estimation accuracy. Then the proposed estimator is constructed so as to absorb as many as possible computable part in the FIM. Using this plan gives the following two features. First, it is that the pseudo regression vector φ(
t-1) instead of the true regression vector φ(
t-1) is related to the filtered estimate of the state vector in a state space representation of ARMA model. Thus the Kalman Filter (KF) is used to produce the φ(
t-1). Second, it is that the information for the error vector φ(
t-1)=φ(
t-1)-φ(
t-1) is used. This plays the role recovering the loss of information in the substitution to φ(
t-1) by φ(
t-1). The resultant estimator is given in the form of linkage with KF. Numerical experiment indicates that convergence rate or the estimation accuracy of parameter estimates is much more improved compared with standard RPLR method.
As an application of the proposed method to real data processing, the modeling on a day evolution of some medical time-series data is dealt. The feature of this data is that the data number is scant. Thus the convergence rate of the estimator must be acclerated to obtain the more accurate estimates. In such the case, the proposed method is useful.
View full abstract
-
Katsuhiko YAMADA, Kazuo TSUCHIYA
1992 Volume 28 Issue 7 Pages
864-871
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
The subject of this paper is a space robot that has a manipulator arm mounted on a satellite body. When the space robot works on a target, it needs to exert force to the target. In this paper, we first derive a force control law of the space robot in case where the space robot and the target are flying freely. Then we apply the force control law to a control scheme after the space robot captures the target. The purpose of the control scheme is to decrease the relative velocity of the target to the space robot without yielding excessive force. The control scheme consists of 2 steps: a step to obtain the desired force and torque exerted on the target from the manipulator hand and a step to obtain the manipulator joint torques to realize the desired force and torque by the force control law. The obtained control scheme is simple, and simulation results show that the control scheme is effective to decrease the relative velocity of the target.
View full abstract
-
Toshimitsu USHIO
1992 Volume 28 Issue 7 Pages
872-878
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This paper introduces a concept of an augmented language of a control specification in discrete-event systems. This language is at least legal in the sense that a plant can not generate illegal strings. We investigate its properties related to controllability of a control specification, and propose an algorithm for computation of the supremal controllable sublanguage, which is based upon an augmented language and a mapping over a family of languages. We show criteria of its finite convergence, which does not require that both a plant and a control specification are described by regular languages, but that the augmented language is regular and “enough large”. We also propose an algorithm based upon an automaton which generates an augmented language in the case that a control specification is closed, and show that its complexity is polynomial.
View full abstract
-
Min LU, Kiyotaka SHIMIZU
1992 Volume 28 Issue 7 Pages
879-886
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
A global optimization technique is proposed for solving a nonlinear programming problem with equality and inequality constraints by using the concept of exterior penalty method. The original problem is transformed into an auxiliary problem which is an optimization problem with inequality constraints only. It is proved that the global optimal solution can be obtained as an accumulation point of a sequence of solutions to the auxiliary problems, as the penalty parameter goes to the infinite. Under appropriate assumptions, we show that the auxiliary problem can be equivalently transformed to a concave program for which we can find a global optimal solution.
It is then applied to solve the bilevel programming problem. We replace the low-level problem by its Kuhn-Tucker conditions, and obtain a nonlinear programming problem with equality and inequality constraints. Then, by applying the above technique, a global optimal solution to the bilevel programming problem is found in the case of convex quadratic lower-level problem.
View full abstract
-
Manabu KOTANI, Haruya MATSUMOTO, Makoto MAKINO, Toshihide KANAGAWA
1992 Volume 28 Issue 7 Pages
887-895
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
We describe the performance of the Auto-association by Neural Network (ANN) for the data compression of the acoustic signal. The network is composed of the multilayer perceptron in which the unit number of the hidden layer is smaller than the others. The error back propagation method is applied for the learning algorithm and the target patterns for the output layer are same as the input patterns.
We, here, apply the ANN to the acoustic diagnosis for the compressor. The purpose of the acoustic diagnosis is to discriminate eight types of faults in the compressor's two elements, the valve plate and the valve spring. We first examine the architecture of the network appropriate for the acoustic diagnosis; it includes the selection of the form of the activation function in the output layer and the number of hidden layers. The relations between the compressed information in the hidden layer and the input original information are also examined, and the comparison of the discrimination accuracy are made for the compressed information, the traditional cepstral coefficients and the filterbank's outputs.
We, then, apply another multi-layered neural network, whose input signal is the compressed information, to evaluate the ability of the data compression in the ANN. The discrimination accuracy is calculated from the degree at which the unit having the maximum output value among the output unit corresponds to the input pattern (experimental condition).
As the result, the structure for the ANN is determined as that the activation function in the output layer is the sigmoid one and the network has one hidden layer. The compressed information is related to the degree of the variance of the input information in each input unit. The discrimination accuracy with the compressed information is 91.1% which is better than that with the traditional cepstral coefficients or the fiterbank's outputs.
View full abstract
-
Koichi MIZUKAMI, Hua XU
1992 Volume 28 Issue 7 Pages
896-898
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This note consider open-loop linear-quadratic Nash games for continuous descriptor systems. We show that there exist no impulse-free open-loop Nash equilibrium solutions if the system contains impulsive modes. But, for its special case of zero-sum open-loop Nash game, a unique impulse-free equilibrium solution may exist if some controllability-observability conditions are satisfied.
View full abstract
-
Shoichi NAKAYAMA, Shin-ichi HORIKAWA, Takeshi FURUHASHI, Yoshiki UCHIK ...
1992 Volume 28 Issue 7 Pages
899-901
Published: July 31, 1992
Released on J-STAGE: March 27, 2009
JOURNAL
FREE ACCESS
This paper presents a knowledge acquisition method using a fuzzy neural network (FNN). The FNN acquires control strategies from operator's manipulating data.
View full abstract