We propose a decomposition method for jobshop scheduling problems by timed automata. Jobshop problem is represented as an optimal state transition problem for timed automata with the parallel composition of submodels composed of jobs and resources in which the latter can be idle or busy. A new decomposition algorithm is proposed to solve the problem. The validity of the proposed decomposition algorithm is confirmed by satisfying the decomposable condition.

View full abstract

We propose a column generation with cut generation for flowshop scheduling problems. In the proposed method, constraints for eliminating some of violation for machine capacity constraints are imposed to the pricing problems for arbitrarily selected two jobs. We show that lower bound can be strengthened by the cut generation without significantly increasing total computing time. Computational experiments show that the proposed method can derive solutions with smaller duality gap compared with those of the conventional column generation.

View full abstract

[in Japanese]

View full abstract

In this research, a method which translates experts' knowledge into action-less rule sets is examined by being applied to a static job shop scheduling problem. The results indicate that modifying the design of the unit-rule used in rule sets to involve characteristics which seem to be prominent in operations with longest completion times can decrease such completion times.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

In this paper, we propose rule induction methods based on rough set theory for decision tables with missing values. Several rule induction methods for decision tables with missing values have been proposed, however, those are based on the knowledge discovery approach. While, the proposed methods are based on the machine learning approach. The proposed methods are based on variable precision rough sets, which can admit error in decision tables. We examine classification accuracy of the proposed methods by numerical experiments.

View full abstract

[in Japanese]

View full abstract

SVM (Support Vector Machine) proposed by Vladimir Vapnik is one of the most widely used learning algoirithm, which uses kernel function to create non-linear classifiers. Kernel functions can be constructed from simple kernel by applying some constructing rules. We therefore propose a method in which genetic programming searches better kernel for solving a problem.

View full abstract

[in Japanese]

View full abstract

In order to measure the purchase demands of accessories, the questionnaire which has basic attributes and attributes on the sense of value was carried out online. The dates are collected approximately 500 samples. In this research, we construct and investigate this model by Bayesian Network whose causal relationship is sequentially chained by (1) the characteristic of customer, (2) the purchase budget and (3) the accessory type.

View full abstract

[in Japanese]

View full abstract

A technique for analytically estimating a relaxed solution for k-Means clustering was proposed based on a PCA-guided manner. In the technique, however, the derived cluster indicator is a rotated solution, in which the rotation matrix cannot be explicitly estimated. This paper introduces a technique estimating rotation matrix by Procrustean transformation of principal component scores in order to select the optimal solution from multiple solutions derived by Fuzzy c-Means, and proposes a cluster validation method considering the deviation between the fuzzy clustering solution and the re-constructed membership indicator matrix, which implies the plausibility of the fuzzy clustering solution from the view point of information summarization.

View full abstract

We have proposed a fuzzy model of a community simulator for behavior analysis in a virtual theme park and demonstrated that a situation where each guest tends to enjoy riding various attractions with concomitant reduce of waiting times can arise by using the strategy based on our proposed fuzzy reasoning model. In this paper, we consider evaluation of the agent models from the view point of the degree that looks like the users. In order to intensify the affinity for agents, satisfaction level is represented by character pictures.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

The data-driven PID control scheme has been proposed for nonlinear process systems by the authors. The effectiveness of the proposed control scheme is examined by employing for an injection molding process.

View full abstract

The control performance of the self-tuning PID control method which is based on on-line identification is depended on the accuracy of the mathematical model. In this work, the re-identification and the re-tuning of the controller are only carried out when the accuracy of the mathematical model is deteriorated.

View full abstract

This paper deals with a design problem of an Anti-Windup adaptive PID control system for SISO systems with a model based parallel feedforward compensator (PFC). In this method, the PFC is designed based on a time-varying ASPR model and approximate system model in order to alleviate windup phenomena which are caused by the saturated input.

View full abstract

[in Japanese]

View full abstract

In this paper, to achieve good handling property even for large variation of vehicle dynamics, a design method for a desired combined vehicle is shown, and then, an adaptive steering controller is developed so that the actual vehicle tracks the desired vehicle. Carrying on numerical simulation, it is shown that the developed adaptive steering controller is very useful for the driver-combined-vehicles system.

View full abstract

[in Japanese]

View full abstract

In this paper, we propose an algorithm of adaptive control scheme, for systems which is unknown, affine and only output is measurable, which converges to the approximate optimal state feedback control solution. In this algorithm, we combine two method. For the first method, we use nonlinear observer which do not use system dynamics. Secondly, we extend adaptive optimal control method which is called Policy Iteration.

View full abstract

Financial data in time series has tendency of dispersion. In general, determination of initial condition plays crucial role since parameters of GARCH model, in which dispersion with condition is symmetry, are estimated by law of Newton-Raphson which depends on the initial condition. In this paper, based on GJR(Glosten/Jagannathan/Runkle) model, which can show nonsymmetrical dispersion with condition, we propose volatility Identification based on GJR model by local and global searches and verify the effectiveness by simulation.

View full abstract

Tele-operation of robot with flexible arm that is able to reduce burden of astronaut is expected. The robot has some problems; apex of arm is in a vibrating manner when stiffness of arm is low value, force of environment could not be felt on labor assistant. In this paper, Vibration suppression control of haptic system based on load disturbance observer is proposed. The proposed method realizes to suppress the vibration of the arm and to feed sense of force back to operator.

View full abstract

Nonlinear optimal regulator problems reduce to the Hamilton-Jacobi equation, which is difficult to solve in general. However, a certain class of problems can be solved by finding a suitable involutive maximal ideal. In this paper, a sufficient condition is shown for nonlinear optimal regulator problems of multiple-input multidimensional systems to be solved with an involutive maximal ideal, and an example of an explicit solution is given.

View full abstract

A stabilization problem for a quantum system is discussed. The states of the system have been described with 1/2 spin variables, and in discrete time domain. Firstly, we introduce the stochastic equation that the system obeys and then, derive an approximated equation for that stochastic dynamics. Secondly, some local stabilities for the approximated and non-approximated target spin systems are analyzed using stochastic Lyapunov function. In the last, we demonstrate those stabilities with numerical simulations.

View full abstract

[in Japanese]

View full abstract

In this paper, we consider the admissibility problem of descriptor systems. By using Lyapunov function we show that certain property of the derivative of the system guarantees the admissibility of problems. Then, we propose a necessary and sufficient condition for the existence of the system to guarantee such admissibility. Finally, we present some numerical examples to illustrate the effectiveness of the obtained results.

View full abstract

The descriptor equation is more useful than the state equation because it can express the wider class of system. However, there are not many researches about the design method of the adaptive control for descriptor systems. In this paper, the design method of the adaptive state feedback control for the asymptotic stabilization problem of multi-input linear descriptor systems is proposed.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

This paper is concerned with a method for sizing pipe wall thinning using guided wave. First an inspection mechanism is described by linear elastic systems in circular cylindrical coordinates. Secondly, two types of excitation forces driven by PZT and by MsS are considered in the numerical experiments. Finally,a computational method for sizing pipe wall thinning is tested with simulation data.

View full abstract

In human behavior, there are some classes of motion patterns, such as walking or running. At the same time, there are motion patterns peculiar to each individual. We suppose that human behavior has a structure consists of similarities not rely on each individual and differences depend on each individual. In this paper, a method for extracting similarities and differences in human behavior using singular value decomposition is proposed, and is applied to time-series data of posture angles of each part of human body in walking and running measured by gyro and magnetic sensors. Experimental results illustrate that similarities and differences of human behaviors are extracted by this method and that human behavior is thought to have a structure consists of them.

View full abstract

In this paper, a method of collision detection for parallel-structured two-link flexible manipulators subject to distributed random disturbance based on the unscented Kalman filter is presented. The manipulator is modeled by the nonlinear stochastic system with Gaussian disturbance. In order to detect when collisions occur, the innovation process of the unscented Kalman filter is used. The performance of the proposed collision detection algorithm is demonstrated by several numerical simulations.

View full abstract

[in Japanese]

View full abstract

A compensation approach of doppler distortion for a multi-way speaker system is proposed and its advantages are verified. A multi-way speaker makes a couple of signal-dependent variable time delays since the compensator using system inversion, a method of making an inverse system, is used to cancel them.

View full abstract

MIDI is a standard protocol for saving and playing music. But a conversion technique from an acoustic signal to MIDI has not been established. This paper proposes a system that can automatically convert an acoustic signal to MIDI data consisting of pitch, pitch bend, and pressure level. The proposed system uses parallel connected adaptive INFs based on zero crossing rate. Since the adaptive INF can pursue a pitch of an acoustic signal, MIDI data can be obtained from the coefficient and output power of the INF. Simulation results show that the proposed system can extract MIDI data from a real acoustic signal.

View full abstract

Many of composition support software marketed now hold the faults that neither individuality nor the peculiar characteristic of the user who uses the software are demonstrated easily enough. We have already proposed the algorithm that offers the sample melody that becomes the hint of the composition using the linear Markov process to improve these faults. We confirmed the composition in which individuality and the peculiar characteristic of the user of this software were reflected by some experiments was able to be done. However, this algorithm had the fault that the note was consecutive more than the necessity. Then, we solved this problem by adding the function to relocate the following note when the note was consecutive more than the necessity. In addition, we thought the length of the producing note was changed by using the prior probability. As a result, the algorithm where abundant sample melodies were able to be produced even by little data was completed.

View full abstract

A method of processing tasks by cooperating with each other between devices including mobile robots installed in a smart space is proposed. The method is that a task divided into several stages is assigned to appropriate one of a group of devices and the top stage is processed on the device, and the rest of stages are handed over another device optimized in order. By repeating these processes, the allotted processes of a task that is occupied with a specific PC can be divided by multiple devices. Thus, this leads to improve the load concentrated on a particular device.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

To operate robotic arm of the International Space Station, it is essentially important to achieve collision avoidance. Since the shape of the robotic arm of the International Space Station that like the robot, has only minimum degree of freedom (DOF) is uniquely decided, the robot cannot avoid an obstacle once the end effecter position determined. But if we admit flexibility in a few DOFs except operators intentions, the DOF to realize collision avoidance can derived. In this paper, an algorithm to achieve collision avoidance using flexibility in a few DOFs except operators intentions is introduced.

View full abstract

There is a concept to utilize module type redundant manipulators for Human support in manned space vehicle. Because the module type redundant manipulators have a lot of redundant degree of freedom and work space in manned space vehicle is limited - the modules need to decide their joint angles in autonomous and distributed manner according to the spatial constraints. In this paper a decentralized autonomous algorithm to decide shape of module type redundant manipulators in the limited space is introduce.

View full abstract

[in Japanese]

View full abstract