「制御系設計における適応・学習・同定・モデリングの新展開」—特集号によせて—　水本 郁朗

抄録全体を表示

The proposed GMVC can be applied to the time varying system. The GMVC has the tracking performance against the reference change and a load disturbance in time varying system.  Turning performance due to ship's steering is different from each steering angle. Therefore, the turning performance due to steering is time varying system. First, this study identifies the ship's maneuverability, which is expressed as Auto-Regressive eXogeneous (ARX) model, by each steering angle. Second, time varying parameters of ARX model are verified. Finally, this study designs a control system for the time varying steering system. The angular velocity of turning is controlled by applying Generalized Minimum Variance Control (GMVC) for time varying system. Especially, GMVC is arranged the polynomial's multiply for the time varying steering system.  The proposed GMVC can be applied to the ship's steering of time varying system. The GMVC can stable steer the ship when a gust of wind is received.  

抄録全体を表示

This paper deals with a design problem of an adaptive PID control with an adaptive Neural Network (NN) feedforward control input for discrete-time systems with a parallel feedforward compensator (PFC). The stability of the designed adaptive PID control system is analyzed provided that the augmented controlled system with a PFC is ASPR. The effectiveness of the adaptive PID control method will be confirmed through numerical simulations for a discrete-time system.

抄録全体を表示

This paper considers the compensation of flow disturbance, which is the variation of air pressure supplied by the air compressor of pneumatic active anti-vibration apparatuses (AVAs). Since the air pressure varies periodically, repetitive control is utilized for the suppression of the vibration induced by the flow disturbance. As the constant gain of the repetitive controller increases, disturbance compensation is improved further. On the other hand, when the gain becomes large, transmissibility increases in the frequency region including the mechanical resonance of AVAs. It leads to the performance degradation of the vibration isolation. In order to overcome the trade-off between the disturbance compensation and the transmissibility, a notch filter is employed instead of the constant gain in the repetitive controller. The center frequency of the filter is set to the frequency region including the mechanical resonance of AVAs so as to avoid the increase of the transmissibility. By using the proposed method, the performance of disturbance compensation can be improved without increasing the transmissibility. Furthermore, effects of parameters for the notch filter are investigated. Experimental and simulation results show that the performance of flow disturbance suppression and the transmissibility of the AVA are almost same regardless of parameters in the notch filter.

抄録全体を表示

This paper proposes a new adaptive trajectory controller in order to realize a good transient performance of trajectory control system for a 2 degree of freedom planer rigid robot arm. This is an improved version of the controller proposed in literature⁽²⁾ and consists of a new friction compensator and a dynamic certainty equivalent adaptive controller. The former does not use any friction models. It is designed as a nonlinear PI controller which directly compensate friction torque. The latter is designed so as to be robust to load change. Then, adjustable parameters are tuned by a dead zone adaptation law which is robust to the existence of bounded friction compensation error. The effectiveness of the proposed controller is verified by a simple analysis and numerical simulation results.

抄録全体を表示

This study discusses a design method to determine one of the design parameters in generalized minimum variance control (GMVC) law from the viewpoint of the control assessment. In this paper, a PI control law is designed based on GMVC. In addition, the control performance of the designed controller is evaluated by using a minimum variance index, and the design parameter in GMVC is automatically changed such that the control performance is optimized. In the proposed method, unknown plant parameters are estimated, and the PI control law based on GMVC is updated using estimated parameters. Finally, the effectiveness of the proposed method is shown through an experiment.

抄録全体を表示

This paper proposes a system identification method for single-input single-output linear time invariant continuous systems only using one-shot closed-loop transient data. The method consists of two methods based on an integral-equation and a direct least-squares approach, respectively, and the former is used as a preprocessing step for the latter. The validity of our proposed method is shown by numerical examples in which the continuous-time transfer functions of a non-minimum phase plant and an unstable plant are identified with the method and their evaluation is achieved.

抄録全体を表示

In this paper, a nonlinear control system considering the uncertainties of the backlash of nonlinear plants is proposed. The uncertainties of the backlash of nonlinear plants are included in thresholds of backlash. Firstly, a backlash model with uncertain thresholds is proposed. Secondly, a robust nonlinear control system is designed. Finally, the effectiveness of the proposed system is confirmed by simulations.

抄録全体を表示

PID control schemes have been widely used in most process control systems for a long time. However, it is a very important problem how to choose suitable PID parameters. Because these parameters give a great influence on the control performance. Especially, it is difficult to tune these parameters for time-varying systems. In this paper, design scheme of a performance adaptive controller is proposed for such systems, which is based on the IMC tuning method. It is well known that the IMC tuning method includes a user-specified parameter. The control performance can be suitably improved by using the user-specified parameter.

抄録全体を表示

Model reference adaptive control systems (MRACS) is the one of the useful systems to control the plant containing unknown variable parameters. However, to design MRACS, it has been needed to satisfy model matching condition. In this paper, to improve the response in the case unsatisfying model matching condition, we design an adaptive law containing fractional order integrator in adaptive law, and prove the system's Lyapnov stability.

抄録全体を表示

In many industrial systems, the multivariable system which has mutual interference has to be dealt with. Therefore, decoupling the multivariable system is an important issue to construct a control system. Although various decoupling methods are proposed from the former, the present condition is that industrial application is not carried out positively. In such a background, the decoupling method in which an inverted decoupler was used, attracts attention due to the handling being easy as compared to the conventional method in recent years. That design approach has a desirable design for the continuous-time system. However, from a practical viewpoint, the design approach in a discrete-time system is desirable. In this paper, an inverted decoupler in a discrete-time system is constructed, and the design method of the multi-loop PID controller to the decoupling system is considered. Specifically, self-tuning control is constructed, whose decoupler and controller are simultaneously adjusted in an on-line manner. The effectiveness of this proposed method is shown by a numerical simulation.

抄録全体を表示

This paper explores safety assessment of self-tuning generalized minimum variance control (GMVC) system by strong stability rate. Strong stability rate is one of the concepts of plant safety and is derived by using coprime factorization approach. Its value is defined by open-loop gain of controlled system through Youla-Kucera parametrization. GMVC is extended by using coprime factorization in this paper. Moreover the extended GMVC is designed so as to make the open-loop gain be closer to the closed-loop gain by new design parameter. In the case that the open-loop gain is closer to the closed-loop one, strong stability rate assesses that the controlled system maintains higher safety in the meaning that the open-loop output will be closer to the closed-loop output even if the feedback signal becomes zero by accident. In this paper numerical example for the self-tuning controller is given in order to verify the validity of the proposed assessment.

抄録全体を表示

The paper gives a controller tuning method using Fictitious Reference Iterative Tuning (FRIT) approaches based on the frequency domain. The proposed method obtains controller parameters that optimize a performance index, which evaluates the difference between the closed loop output and the reference model output in the frequency domain. The performance index is calculated using the Fourier transformation of one-shot experimental input-output data. The paper shows that band-path filters play an important role in the Fourier transformation of the one-shot input-output data, and the optimal pre-filter, which well approximates the performance index to the original performance index, can be easily calculated in the frequency domain. The optimal pre-filter contributes to deriving a better controller parameter improving the closed loop response. The effectiveness of the proposed method is illustrated through numerical examples for PID gains tuning.

抄録全体を表示

The paper proposes a direct control parameter tuning method using a generalized minimum variance (GMV) evaluation in regulatory control. In the method the performance index is calculated using the squared difference between an estimated variance of generalized output and a desired generalized output variance would be minimized. A one-shot closed loop experimental input and output data by an initial controller generate the estimated variance of generalized output instead of using a plant model. Therefore, the proposed method does not require an extra plant test for the identification of the plant model and/or control parameter tuning. Furthermore, since the GMV evaluation consisting of weighted output and control signal is employed, excessive high gain control parameters can be avoided. The feature can be expected to improve stability of the closed loop system. The efficiency of the proposed method is demonstrated from a numerical example.

抄録全体を表示

This paper presents a new design method for a self-repairing nonlinear control system with an adaptive tuner for unknown plants with sensor failures. In this method, a nonlinear detection filter is utilized to detect the sensor failure exactly and early. Furthermore, by introducing the adaptive tuner for high-gain feedback, the regulated output can asymptotically enter a prescribed ball with an arbitrarily small radius. This paper theoretically shows that both self-repairing (switching sensors) and adaptive regulation can be achieved. Also, to confirm the effectiveness of the proposed control system, numerical simulation is explored.

抄録全体を表示

This paper studies the flight control of a twin-rotor type model helicopter. When the model of the helicopter is linearized using the first-order approximate method, the attitude angular and the move angular of the helicopter are decoupled. Therefore the linearized system is uncontrollable. This paper proposes a design method for controlling the helicopter using interference. Finally, the effectiveness of the proposed method is shown through the simulation.

抄録全体を表示

This paper proposes a new method for designing a multi-rate control system, in which the sampling interval of the plant output is an integer multiple of the hold interval of the control input. The objective of this study is to enhance a noise response without changing an existing closed-loop transfer function. To this end, an existing control law is redesigned independent of the closed-loop transfer function. In this study, the noise response is designed independent of the sampled reference response, and the effectiveness of the proposed method is shown.

抄録全体を表示

This paper proposes the switching control method based on the current control performance. The current control performance is obtained in an online manner over a user-specified time-window with some overlap. 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

「インテリジェント・システム」—特集号によせて—　相吉 英太郎

抄録全体を表示

The surplus power that will be caused by large penetration of PV in power system can affect the stable operation of power system and degrade the reliability of power supply. In order to overcome the above problem, cooperative operation with customer's appliances, such as heat pump water heater (HPWH) and battery, is proposed. This paper shows a study case of utilizing surplus power of Photovoltaic generation (PV) system by using customers' appliances.

抄録全体を表示

Study of heritage database and historical archive become an active area in multidisciplinary fields. The huge and high dimensional content is a sort of big data consisting of various kinds of information elements, and has valuable knowledge and the wisdom of mankind. This study focuses on Kosode Byobu collection which is antique fine art in Edo era. By analyzing high-resolution digital images of them and extracting knowledge structure by machine learning, we apply it for intelligent display and research of the art. This article introduces the background of the project and shows fundamental approach for tackling to retrieving intelligent structure and developing effective display technique.

抄録全体を表示

This paper describes a non-fragile controller design method for uncertain discrete-time systems. Non-fragile controller takes care of the uncertainties in the system parameters and the control gain, which may occur due to posible modeling errors and failure of the actuator. The problem to be addressed is the non-fragile stabilizing and H_∞ control design problem of output feedback controller. Sufficient conditions for the solvability of the stabilizing and H_∞ control design problem for uncertain discrete-time systems are obtained. Based on such conditions, non-fragile control design methods are proposed via linear matrix inequality (LMI). Finally, a numerical example is given to show the effectiveness of our control design method.

抄録全体を表示

Recently, Particle Swarm Optimization (PSO) algorithm is one of the popular methods for solving optimization problems. In this paper, in order to presume the construction of the solution landscape, we integrate PSO and Response Surface Method (RSM). Moreover, we propose a new method to decompose solution space into several 2-dimensional search space so as to reduce computation complexity. We have confirmed the characteristics and validity of the proposed method by computer simulations.

抄録全体を表示

In recent years, portable computing devices (PCDs) are becoming very popular. To improve the quality of service (QoS) for each individual user, it is necessary to develop application programs that can be aware of the user intention, preference, situation, etc., so that proper services can be recommended at proper timing. We call these kinds of programs awareness agents (A-agents). To satisfy various needs of a user, many A-agents should work together in one PCD. Since the computing resource in a PCD is limited, it is necessary to reduce the implementation costs of the A-agents while preserving their performance. For this purpose, we propose two decision boundary making (DBM) algorithms in this paper. The basic idea of DBM is to generate new training data using given ones to fit the decision boundary (DB) of the given problem, and induce small neural networks (NNs) using the new data. Both algorithms proposed here are simplified versions of the decision boundary learning (DBL) algorithm proposed by us earlier. Using the new algorithms, the cost for data generation can be greatly reduced. Experimental results show that if the new data are generated properly in positions close to the DB, the induced small NNs can perform even better than support vector machines, which are known as the state-of-the-art learning models.

抄録全体を表示

In this paper, we propose a method to diminish the state space explosion problem of a multiagent reinforcement learning context, where each agent needs to observe other agents' states, and previous actions at each step of its learning process. However, both the number of state and action become exponential in the number of agents, leading to enormous amount of computation and very slow learning.  In our method, the agent considers other agents' statuses only when they interfere with one another to reach their goals. Our idea is that each agent starts with its state space which does not include information of others'. Then, they automatically expand and refine their state space when agents detect interference. We adopt the information theory measure of entropy to detect the interference status where agents should take into account the other agents. We demonstrate the advantage of our method over the properties of global convergence in a time efficient manner.  

抄録全体を表示

Recently, applications of multiagent systems are expected from the view point of the parallel and distributed processing of systems. Reinforcement learning attracts attention as an implementing method of multiagent systems. However, there is a problem that the more the number of agents to deal with increases, the slower the speed of learning becomes. To solve this problem, we propose a new reinforcement learning method that can learn quickly and reduce amount of memory. It tries to increase efficiency of the learning on a hunter game by paying attention to partial states of two agents among a large number of agents. In addition, the proposed method employs a switching algorithm and detects automatically a switching time by using a special index called golden cross.

抄録全体を表示

This paper proposes a new reinforcement learning method to construct agents in environments with cyclic reward depending on time. The proposed method consists of two parts: (a) a cyclic action-value function by superposing sinusoidal action-value function in phasor representation and (b) an algorithm to use it. Reinforcement learning is a widely used framework to develop agent which can decide suitable action. It enables the agent to learn suitable action only in stationary environments. Contrast to conventional methods, the proposed reinforcement learning method can be applied to learning in environments with cyclic reward depending on the time. Experimental results show that the proposed method performs much better than conventional methods.

抄録全体を表示

We propose a simple but useful concept of the constrained optimization using evolutionary algorithms. In conventional constraint-handling approaches, it is desired that the objective and constraints should be improved simultaneously. In contrast, in the proposed approach, with an additional use of feasibilization operations to all new solutions, a constrained problem can virtually be recognized as an unconstrained one and the objective function can be improved only in the feasible region. The feasiblization operation is realized by solving a set of nonlinear equations regarding constraints. Nevertheless, in case of possible failure of feasibilization due to limitations of nonlinear solvers, the idea of ε level comparison is applied to evaluate solutions. We also propose two practical algorithms: Feasibilization Particle Swarm Optimization (FPSO) and Feasibilization Differential Evolution (FDE). We show the usefulness of the proposed method by numerical experiments; in particular, both of them perform as well as or better than existing promising methods for engineering design problems.

抄録全体を表示

Recently, a paradigm of optimization method called meta-heuristic (MH) has attracted interest for its applicability. However, the most part of MHs have insufficient mathematical background in their dynamics. Some of MHs adopt pseudo-random numbers generated by a computer in their algorithm so that they are called “stochastic optimization methods”. As for stochastic optimization methods, the positions of search points are able to be regarded as stochastic variables which are distributed according to a certain probability distribution. In this paper, we perform an analysis of search points distribution for DE dynamics based on maximum entropy method.

抄録全体を表示

Multidisciplinary design optimization (MDO) using design of experiments (DoE) and response surface method (RSM) is one of the effective methods for reducing weight of car body structure. However, large amount of information obtained from the optimization makes the optimization itself not useful. This problem is caused by large number of variables of vehicle model. This paper proposes a trade-off analysis process using sensitivity analysis, multi-objective optimization, hierarchical cluster analysis and parallel coordinate plot for analyzing optimization information. As a result, it was clarified very important parts for reducing body weight. Accordingly, even the designers without the expert knowledge may easily be able to make a knowledge discovery for car-body weight reduction.

抄録全体を表示

Evolutionary swarm robotics (ESR) is an artificial evolution approach to generating meaningful swarm behavior in multi-robot systems which typically consist of many homogenous autonomous robots in which the same robot controllers designed with evolving artificial neural networks are employed. Historically speaking, this approach has been thought to be a promising approach for swarm robotics systems (SRS), because the swarm behavior is an emergent phenomenon caused by many local interactions among autonomous robots and it is very hard to give a program to each robot for appropriate swarm behavior in advance. However, its realization is considered to be impractical even for a simulated SRS because the artificial evolution requires a very large computational cost. In this paper, in order to overcome this computational cost problem, a novel implementation method, i.e., the parallel problem solving using graphics processing units (GPUs) and OpenMP on a multi-core CPU, is introduced. The efficiency of the proposed method is demonstrated with the food-foraging problem with an evolving SRS.

抄録全体を表示

We develop a tool that finds the optimal system configuration and the optimal operation rules of a hybrid water heater system. A hybrid water heater system consists of three elements, a heat pump water heater, a boiler, and a storage tank. The tool finds the optimal output of a heat pump water heater, the optimal output of a boiler, and the optimal capacity of a storage tank. Furthermore the tool finds the optimal operation parameters that determine startup and shutdown of a heat pump water heater and a boiler.

抄録全体を表示

We propose a new Particle Swarm Optimization (PSO) algorithm which utilizes the information about the distribution of personal bests (pbests). Basically, it applies the standard PSO; however, when the global best (gbest) approaches an optimal solution, its attracting region is estimated by using the distribution information. If particles gather around the gbest, the swarm is divided into two sub-swarms: (a) the local search sub-swarm, which keeps searching for the local solution by using the standard PSO; and, (b) the other solutions search sub-swarm, which moves particles to different solutions by using a modified PSO. When the local search is completed, the standard PSO is appled to all the particles again to trigger the estimation of the attracting region of another optimal solution. Additional resetting of particles in several situations is also applied to keep the diversity of global search. We show the usefulness of the proposed method by numerical experiments.

抄録全体を表示

It is desired to make the replication portfolio when a benchmark portfolio has delivered good returns. However, the portfolio replication problem is one of equality constrained indeterminate problems. We cannot find the same proportion-weighted combination as the benchmark portfolio by minimizing the evaluation values of solutions even if we use any evolutionary algorithm. We have an important problem that minimizing the evaluation value does not depend on improving the solution. In order to solve this problem, we propose a new model which removes the equality constraint in the optimization problem in this paper. Our model transforms the equality constrained search space to the unconstrained search space for the portfolio replication problems. In the numerical experiments, we show that the evolutionary algorithms can generate the good solutions by minimizing the evaluation values in the unconstrained search space obtained by our model.

抄録全体を表示

As the growing penetration of HEMS (Home Energy Management System) and smart meters, systems to dynamically process demand data and then to automatically provide information on energy conservation are starting to receive wide attention, especially in the U. S. Not only to utilize such systems for customers in Japan, but also to maximally enhance the effect, it is essential to consider the particular problem in Japan that reducing peak-time demand of electric system is especially expected, alongside to consider how real human beings react on provided information. We propose a method to automatically generate the reports on energy conservation tips to ‘Nudge' customer's energy conserving behaviors, based on each customer's electricity demand data. The reports feature the aspects of real human behaviors, including search cost, social norms, and expression to emphasize nuances of the information, to mainly aim at reducing electric system's peak-time demand. The paper includes the preliminary analysis results of the effect of reports generated by the proposed method, and further necessary points to be analyzed from the viewpoint of actual behavioral change are also discussed.

抄録全体を表示

The future of accelerator-based (Acc-based) neutron irradiation system (NIS) for boron neutron capture therapy (BNCT) using liquid lithium target was studied especially for the usage of neutrons from ⁷Li(p,n)⁷Be near threshold reactions. The next generation of NIS for BNCT is envisioned to have a non-invasive dose monitoring system during clinical BNCT performed at a downtown hospital. A liquid lithium film flow for the neutron producing target for BNCT has been established. The combination of a long-life neutron producing target such as a liquid lithium target and a stable proton accelerator such as RFQ is evaluated as a promising candidate for this future system. A new design standard for an Acc-based NIS for BNCT derived from the viewpoints of secondary cancer induced by BNCT, was proposed.

抄録全体を表示

Norepinephrine (NE) has a great important role in antiepileptic effects in vagus nerve stimulation and neuromodulation. There exists knowledge about NE effects on organ and neuron, however little is known about ones on neuronal network. Here we evaluated the NE effects on cultured neuronal network on microelectrode array (MEA) with analysis based on spikes, electrode bursts which are bursts recorded in a single electrode, network bursts, and Fourier spectrum. Multiple NE administrations show significant decrease in number of spikes, electrode bursts, and network bursts and other indices of those activities. Furthermore, average spike rate and normalized Fourier spectrum show that the administrations decrease the baseline of activity and emphasize periodic activity. These results suggest that NE has antiepileptic effects by decreasing synchronized activity and neuromodulation by emphasis on periodic activity.

抄録全体を表示

＊投稿手引の改正に伴い，2013年10月1日以降に新規に投稿され，その後掲載決定となった論文および資料には，Extended Summaryがありません。

抄録全体を表示