本記事に「抄録」はありません。

抄録全体を表示

In this paper, we give a tutorial review of fictitious reference iterative tuning (which is abbreviated as FRIT in the following). This tuning method enables us to obtain the optimal parameter with only one-shot experimental data of the closed loop and performing an off-line nonlinear optimization. Thus, this method is useful from the practical points of view in the case in which it is difficult to take a time and cost for tuning of a parameterized controller. In this expository article, we give the basic idea of FRIT and how to apply this method to practical problems. Particularly, we focus on the problem how to tune a controller parameter so as to let the closed loop track for desired responses. Moreover, we mention the future developments of this tuning method.

抄録全体を表示

This review paper shows a PID gains tuning method from one-shot experimental data generated by test signal added at input signal in an off-line manner, so that the output signal could follow the prescribed reference model output. We call the method a “disturbance attenuation FRIT method” because the test signal added at input signal is a benchmark signal evaluating disturbance attenuation property. The experimental result for helicopter attitude control model is demonstrated to show the effectiveness of the disturbance attenuation FRIT method.

抄録全体を表示

In the challenge to manufacture high quality products for less, it is necessary to regularly monitor performance of control loops that regulate the quality variables of interests. This paper introduces two methods of performance-driven PID control systems whose PID parameters are re-tuned based on the controller performance assessment. The re-tuning is based on the self-tuning control mechanism. The effectiveness of these schemes are demonstrated by the experimental studies.

抄録全体を表示

Authors proposed the estimation method combining k-means algorithm and NN for evaluating massage. However, this estimation method has a problem that discrimination ratio is decreased to new user. There are two causes of this problem. One is that generalization of NN is bad. Another one is that clustering result by k-means algorithm has not high correlation coefficient in a class. Then, this research proposes k-means algorithm according to correlation coefficient and incremental learning for NN. The proposed k-means algorithm is method included evaluation function based on correlation coefficient. Incremental learning is method that NN is learned by new data and initialized weight based on the existing data. The effect of proposed methods are verified by estimation result using EEG data when testee is given massage.

抄録全体を表示

This paper deals with a design problem of an adaptive output feedback control for discrete-time systems with a parallel feedforward compensator (PFC) which is designed for making the augmented controlled system ASPR. A PFC design scheme by a FRIT approach with only using an input/output experimental data set will be proposed for discrete-time systems in order to design an adaptive output feedback control system. Furthermore, the effectiveness of the proposed PFC design method will be confirmed through numerical simulations by designing adaptive control system with adaptive NN (Neural Network) for an uncertain discrete-time system.

抄録全体を表示

In this paper, we propose a direct tuning method of linear multivariable controllers based on frequency response measurements. Prior to the controller tuning, the plant is stimulated by some appropriate block-pulse excitation signals, and the frequency response measurements are obtained by DFTs of the input-output data. Then a linearly parameterized controller is obtained by minimizing a frequency-shaped model reference control criterion at selected frequency points in the frequency-domain. Compared to some existing time-domain data-based controller tuning methods, the method usually does not need iterative experiments. And the control performance seems to be more transparent, owing to the visualized frequency-domain data. The performance of the proposed method is shown through extensive simulation studies.

抄録全体を表示

In this paper, an application method of Fictitious Reference Iterative Tuning (FRIT), which has been developed for controller gain tuning for single-input single-output systems, to state feedback gain tuning for single-input multivariable systems is proposed. Transient response data of a single-input multivariable plant obtained under closed-loop operation is used for model matching by the FRIT in time domain. The data is also used in frequency domain to estimate the stability and to improve the control performance of the closed-loop system with the state feedback gain tuned by the method. The method is applied to a state feedback control system for an inverted pendulum with an inertia rotor and its usefulness is illustrated through experiments.

抄録全体を表示

Almost real-world processes are represented as nonlinear systems. Therefore, it is important to consider control schemes to cope with such systems. In this paper, a control method for nonlinear systems is newly proposed. Some local linear models on typical equilibrium points are first designed, followed by linear controllers corresponding to these models. The distances between the query and these local models are calculated, and the weights are computed in proportion to the distances. These weights are put for local controllers, and the controller corresponding to the query can be designed. According to the proposed scheme, the good control performance can be easily obtained. The effectiveness of the proposed control scheme is shown by some simulation examples.

抄録全体を表示

The decision of sampling and hold intervals is an important issue in the design of sampled-data control systems, in which a continuous-time plant is controlled using a digital computer in discrete time, and the selection of intervals greatly influences the control quality. However, it is not easy to set these intervals arbitrarily because of hardware constraints. This paper proposes a new method for the design of a multi-rate single-input single-output sampled-data control system under the constraint that the sampling interval of the plant output in continuous time is an integer multiple of the hold interval of the control input, which is updated in discrete time. In the design of the multi-rate control system, the intersample plant output can possibly oscillate even if the sampled plant output converges to its reference input. In conventional design methods, to improve intersample ripples, both the sample and the intersample responses are simultaneously designed, and then, the original sample response might be deteriorated even though the intersample response is improved. In this study, a multi-rate control law, which stabilizes a closed-loop system, is extended on the basis of state-space representation, and the multi-rate control system is redesigned such that the steady-state intersample response is improved independently of the sample response. As a result, the sample response is maintained and the intersample response is improved in the steady state. Numerical examples demonstrate the effectiveness of the proposed method.

抄録全体を表示

In industrial processes, PID control has been applied to a lot of real systems. The control performance strongly depends on PID parameters. Although, some schemes for tuning PID parameters have been proposed, however, these schemes need system parameters which are estimated by system identification in order to calculate PID parameters.
On the other hand, model-free controller design schemes represented by VRFT or FRIT which have received much attention in last few years. These methods can calculate control parameters using closed-loop data and are expected to reduce computational costs.
In this paper, a type of implicit PID controllers using closed-loop data is proposed. According to the proposed method, PID parameters are calculated based on the implicit GMVC. Moreover, the control performance can be suitably adjusted by only user-specified parameter. The effectiveness of the proposed method is numerically and experimentally evaluated.

抄録全体を表示

This paper explores an application of two degree-of-freedom generalized predictive control (GPC) to temperature control of an aluminum plate. Two degree-of-freedom GPC has the feature that its controller reveals an effect of integral compensation only if there is modeling error or disturbance. Therefore performance degradation due to an integral compensation, such as excessive control effort or slow response, can be expected to avoid. In this paper, this feature is defined as two degree-of-freedom system because the characteristics of output response with and without modeling error and disturbance can be designed independently, that is, on one hand the characteristic of output response without modeling error and disturbance is designed by minimization of objective function which includes control input and tracking error, on the other hand the characteristic of output response with modeling error and disturbance is designed by a gain of integral compensator. In order to verify the validity of proposed method, this paper explores numerical simulation and experiment for an aluminum plate thermal process with proposed method.

抄録全体を表示

In this paper, stabilization problem for a class of multi-channel Markov jump stochastic systems with additive gain perturbations is addressed. In order to design the state feedback stabilization controller, a novel control strategy that guarantees the adequate cost bound is established. A sufficient condition for the existence of the controller is derived in term of matrix inequality. In the sequel, for practical plants since it is not always easy to access the transition modes, the mode-independent strategy set is considered. As a result, it is shown that such strategy set can be obtained by solving the linear matrix inequality (LMI). Furthermore, in order to reduce the cost performance, fuzzy control for each input is combined with the static gain as additive gain perturbations. As a result, even if dramatic random parameter changes occur in the system, it is shown that the robust stability and the reduction of the cost can be attained. Finally, a simple numerical example is given to demonstrate the availability of the proposed method.

抄録全体を表示

Switching and ON/OFF controls are effective control techniques for control systems equipped with low-resolution actuators. They can be modeled as control systems that restrict the control input to discrete values. In this paper, a controller design method based on a machine learning technique is discussed. The relation between the current situation (previous input sequence and previous output sequence), applied input, and output evolution is learned on the basis of some machine learning methods. Specifically, different machine learning methods, such as approximate nearest neighbour (ANN) method and support vector machine (SVM) are used in this study. The trained classifier will be a controller that connects current situation and suitable control input that can drive the current output to the desired one. The effectiveness of the proposed method is verified for discrete input systems via some simulations and experiments.

抄録全体を表示

This paper proposes an adaptive PID controller which is driven by current control performance. The calculations of the PID parameters are based on the generalized minimum variance control(GMVC) algorithm. The current control performance is obtained in an online manner over a user-specified time-window with some overlap. The retuning of PID parameters are only carried out when controller performance deteriorates below a user-specified threshold. Experimental evaluations on the voltage control of the DC-DC converter demonstrates the practicality and utility of this idea.

抄録全体を表示

In this paper, a newly modeling system by using Genetic Algorithm (GA) is proposed. The GA is an evolutionary computational method that simulates the mechanisms of heredity or evolution of living things, and it is utilized in optimization and in searching for optimized solutions. Most process systems have nonlinearities, so it is necessary to anticipate exactly such systems. However, it is difficult to make a suitable model for nonlinear systems, because most nonlinear systems have a complex structure. Therefore the newly proposed method of modeling for nonlinear systems uses GA. Then, according to the newly proposed scheme, the optimal structure and parameters of the nonlinear model are automatically generated.

抄録全体を表示

In atmospheric distillation processes, the stabilization of processes is required in order to optimize the crude-oil composition that corresponds to product market conditions. However, the process control systems sometimes fall into unstable states in the case where unexpected disturbances are introduced, and these unusual phenomena have had an undesirable affect on certain products. Furthermore, a useful chemical engineering model has not yet been established for these phenomena. This remains a serious problem in the atmospheric distillation process. This paper describes a new modeling scheme to predict unusual phenomena in the atmospheric distillation process using the GMDH (Group Method of Data Handling) network which is one type of network model. According to the GMDH network, the model structure can be determined systematically. However, the least squares method has been commonly utilized in determining weight coefficients (model parameters). Estimation accuracy is not entirely expected, because the sum of squared errors between the measured values and estimates is evaluated. Therefore, instead of evaluating the sum of squared errors, the sum of absolute value of errors is introduced and the Levenberg-Marquardt method is employed in order to determine model parameters. The effectiveness of the proposed method is evaluated by the foaming prediction in the crude oil switching operation in the atmospheric distillation process.

抄録全体を表示

This paper studies identification in H^∞ based on multiple sets of data in the frequency domain. The traditional H^∞ worst-case identification and the stochastic identification settings are revisited, and a new problem formulation is presented by modifying the H^∞ worst-case identification setting. In the literature, given a single set of data, the limited experimental information subject to finite samples has been focused on, while this paper considers uncertainty modeling given multiple sets of data. It is shown that the problem is solved by means of linear matrix inequalities and subspace identification methods in the frequency domain.

抄録全体を表示

This paper treats a discrete wavelet transform(DWT)-based image watermarking with considering the false positive probability and bit-pattern embedding. We propose an iterative embedding algorithm of watermarking signals which are K sets pseudo-random numbers generated by a secret key. In the detection, K correlations between the watermarked DWT coefficients and watermark signals are computed by using the secret key. L correlations are made available for the judgment of the watermark presence with specified false positive probability, and the other K-L correlations are corresponding to the bit-pattern signal. In the experiment, we show the detection results with specified false positive probability and the bit-pattern recovery, and the comparison of the proposed method against JPEG compression, scaling down and cropping.

抄録全体を表示

Recently, the distortion caused by signal transduction has been significantly reduced because of development of digital signal processing technology, so that sound quality has improved dramatically. However, speaker system located at the end of the sound reproduction has not changed since the basic principle of the invention, the non-linear distortion occurs mostly have been concerned about the deterioration of sound quality. It is important to eliminate this distortion in high fidelity music playback. In order to reduce the nonlinear distortion, we propose a system using disturbance observer and consider in this method.

抄録全体を表示

The difference by the brain activity when writing it down speech the use of the near-infrared spectroscopy is examined in the present study. It is thought that it becomes a help of the communications between the nurse and people requiring long-term care if the identification of the information transmission means can be evaluated and be detected from the brain activity. Because the possibility of the evaluation of result, “Speech”, and “Writing” was shown, it reports.

抄録全体を表示

This paper describes a novel obstacle avoidance control scheme of electric powered wheelchairs for realizing the safe driving in various environments. The “electric powered wheelchair” which generates the driving force by electric motors is expected to be widely used as a mobility support system for elderly people and disabled people; however, the driving performance must be further improved because the number of driving accidents caused by elderly operator's narrow sight and joystick operation errors is increasing. This paper proposes a novel obstacle avoidance control scheme based on fuzzy algorithm to prevent driving accidents. The proposed control system determines the driving direction by fuzzy algorithm based on the information of the joystick operation and distance to obstacles measured by ultrasonic sensors. Fuzzy rules to determine the driving direction are designed surely to avoid passers-by and walls considering the human's intent and driving environments. Some driving experiments on the practical situations show the effectiveness of the proposed control system.

抄録全体を表示

We have developed demodulated direct current voltage-based microwave CT (DC-MCT) which is composed on the basis of chirp-pulse microwave CT. DC-MCT was used time domain measurement technique to determine the shortest propagation path between the transmitting-and receiving-antennas. By using standard signal generator with the low phase noise and phase synchronization of signal source, the reconstructed amplitude-and phase-images can be obtained. A study of high speed imaging, the method of measurement and the influence of overshoot characteristics from the low pass filter were examined, then these images were evaluated for the number of rotary scanning of 50 and 72. In addition, for the purpose of improvement of resolution, pursuing a time domain spectrum which gives the max value of amplitude was showed the possibility to improve receiver dynamic range. At the result, this paper shows usefulness of new method for microwave imaging.

抄録全体を表示

We propose a hand-eye calibration method for 3-D vision sensors by using only a mark. Main advantage of the method is safety and easy to use. It uses only a mark and needs only five observations. The method has two points; 1) A Rotation matrix and a translation vector is independently computed. 2) Each observation is integrated as much as possible. In this paper, we show problem settings, formulation of the method, observation conditions and results of performance evaluation experiments.

抄録全体を表示

In this paper, a new method of motion analysis using attractors in nonlinear dynamical systems is discussed. The attractor is defined as a set of spatially-expanded trajectories of time-series data of a human motion in a state space. Using the attractor representation in the state space, a method of feature extraction from time-series data of human motions is proposed. The time-series data of human motions are captured by wearable inertial sensors. First, a design method of a dynamical system, which encodes time-series data of motions in attractors, is introduced. Next, an example of feature extraction using our approach is demonstrated for a simple upper limb movement. Finally, the physical meaning of the extracted features is discussed. As a result, the extracted features by attractors can describe the characteristics of the human motion, such as posture and quickness, effectively in the spatiotemporal continuity feature space.

抄録全体を表示

When we call the production flow to transition elements in the next step in the process of product manufactured one, the production flow is considered to be displaced in the direction of the unit production density. Density and production, as captured from different perspectives, also said production costs per unit of production. However, it is assumed that contributed to the production cost of manufacturing 100 percent. They may not correspond to the physical propagation conditions after each step of the production density, the equations governing the manufacturing process, which is intended to be represented by a single diffusion equation.
We can also apply the concept of energy levels in statistical mechanics, production density function, in other words, in statistical mechanics “place” that if you use the world of manufacturing and production term. If the free energy in this production (potential) that are consuming the substance is nothing but the entropy production. That is, productivity is defined as the entropy production has to be. Normally, when we increase the number of production units, the product nears completion at year-end number of units completed and will aim to be delivered to the contractor from the turnover order. However, if you stop at any number of units, that will increase production density over time. Thus, the diffusion does not proceed from that would be irreversible. In other words, the congestion will occur in production. This fact and to report the results of analysis based on real data.

抄録全体を表示

The problem of the series type adaptive filter using LMS algorithm is the bias error of the tap gain caused by the noise. In this paper we consider a noise free fast SSCF (Square Sum of Correlation Function) adaptive algorithm with white input signal and its application to the adaptive equalizer. The comparison of the convergence speed to LMS and SSCF adaptive algorithm is presented. The convergence of the adaptive algorithm is guaranteed under the assumption of the uncorrelated property of the noise and the white input signal. Finally, these results of the theoretical consideration are verified by computer simulation.

抄録全体を表示

Human beings have the ability to perceive continuous curves from a binary image that includes discontinuous curves and noises. To achieve this capability, global detection is required. In the past, we have proposed a perceptive curves detection method with accumulating evaluation value of a partial curve. In addition, this paper proposes a method that is highly accuracy by global edge tracing. Experimental result shows that curves by proposed method include no interruption and branching. Its subjective assessment is the highest from other methods.

抄録全体を表示

Robust control theory generally guarantees robustness and stability of the closed-loop system. It however requires a mathematical model of the system to design the control system. It therefore can't often deal with nonlinear systems due to difficulty of modeling of the system. On the other hand, reinforcement learning methods can deal with nonlinear systems without any mathematical model. It however usually doesn't guarantee the stability of the system control. In this paper, we propose a “Real-time Reinforcement Learning Control System (RRLCS)” through combining reinforcement learning to treat unknown nonlinear systems and robust control theory to guarantee the robustness and stability of the system. Moreover, we analyze the stability of the proposed system using H_∞ tracking performance and Lyapunov function. Finally, through the computer simulation for controlling an inverted pendulum system, we show the effectiveness of the proposed method.

抄録全体を表示

We undertake numerous activities in our daily life and for some of those we forget to complete the action as originally intended. Significant aspects while performing most of these actions might be: “pairing of both hands simultaneously” and “short time consumption”. In this work an attempt has been made to recognize those kinds of Paired Activities (PAs), which are easy to forget, and to provide a method to remind about uncompleted PAs. To represent PAs, a study was done on opening and closing of various bottles. A model to define PAs, which simulated the paired behavior of both hands, is proposed, called “Paired Activity Model” (PAM). To recognize PAs using PAM, Paired Activity Recognition Algorithm (PARA) was implemented. Paired motion capturing was done by accelerometers, which were worn by subjects on the wrist areas of both hands. Individual and correlative behavior of both hands was used to recognize exact PA among other activities. Artificial Neural Network (ANN) algorithm was used for data categorization in PARA. ANN significantly outperformed the support vector machine algorithm in real time evaluations. In the user-independent case, PARA achieved recognition rates of 96% for only target PAs and 91% for target PAs undertaken amidst unrelated activities.

抄録全体を表示

This paper proposes some basic methods of content searching with gaining high user utility or user's satisfaction from a lot of information sources scattered in large scale networks. In these methods, searching policy is decided by the estimated user utility gain and/or current user utility for each searching action, and the most favorable combination of an information source and searching content item is selected so as to enlarge the total amount of user utility of whole searching activity. In addition, we evaluate the efficiency of the methods by computer simulations; in which, we give network topology, user utility functions, existing probabilities of searching content items in each information source etc., apply proposed methods and other methods for comparison, and compare summation of their utility gains as an estimative index. As a result, the proposed methods achieve good performance in general and about 2.45 times larger index value than an ordinary searching is observed in an extreme simulation scenario. Furthermore, we discuss pros and cons of the proposed methods in all the simulation scenarios.

抄録全体を表示

This paper proposes a new framework for B to B collaboration process based on a concept of service. Service value, which gives users satisfaction for provided services, depends on the situation, user characteristics, and user objectives in seeking the service. Vargo proposed Service Dominant Logic (SDL), which determines service value according to “value in use”. This concept illustrates the importance of the relationship between the service itself and its situation. This relationship is analogous to electro-magnetic field theory in physics. We developed the concept of service fields to create service value based on an analogy of the electro-magnetic field. By applying this concept to B to B collaboration, a model of service value co-creation in the collaboration can be formulated. Then, the collaboration can be described by 4 steps of KIKI model (Knowledge sharing related to service system, Identification of service field, Knowledge creation for new service idea, Implementation of service idea). As its application to B to B collaboration, the energy saving service business is reported to demonstrate the validity of the proposed collaboration model. This concept can be applied to make a collaboration process effective.

抄録全体を表示

This paper addresses the prioritization problem of parameters for making consensus on the combination of risk-reducing plans. However, the existing method can not prioritize the parameters correctly based on the differences in “preferences”. So, we propose a prioritization method of adjusting parameters by mutual effect analysis. The proposed method defines the change rate of the preference as the mutual effect. Whether the mutual effect works or not is decided by “independent score” with the change rate. The risk-reducing plan which has strong mutual effect is prioritized in descending order of the mutual effect represented as “influence score”. As a result, the proposed method can improve averages of occupancy of parameters to be adjusted in top 5 of the priorities by 30%, and allow the parameter adjustment to be completed in an hour.

抄録全体を表示

To lower power consumption has becomes a worldwide concern. It is also becoming a bigger area in Computer Systems, such as reflected by the growing use of software-as-a-service and cloud computing whose market has increased since 2000, at the same time, the number of data centers that accumulates and manages the computer has increased rapidly. Power consumption at data centers is accounts for a big share of the entire IT power usage, and is still rapidly increasing. This research focuses on the air-conditioning that occupies accounts for the biggest portion of electric power consumption by data centers, and proposes to develop a technique to lower the power consumption by applying the natural cool air and the snow for control temperature and humidity. We verify those effectiveness of this approach by the experiment. Furthermore, we also examine the extent to which energy reduction is possible when a data center is located in Hokkaido.

抄録全体を表示