In this paper, high speed synchronization of two positioning axes under self-tuning fuzzy feedforward control is considered. The system consists of a self-tuning fuzzy feedforward controller(STFFC), a constant gain feedback controller and a synchronizing controller. The fuzzuy rules of the STFFC are adjusted by a learning algorithm so that the tracking error and the synchronizing error are minimized. After the tuning process is comnpleted, the self-tuning fuzzy feedforward controller is the equivalent dynamic inverse of each axis. Experimental results demonstrate the effectiveness of the proposed system for high speed sunchronization.

View full abstract

This paper considers fuzzy facility location problem with rectilibear distance. We formulate the location problem as max-min satisfaction degree type of demand points, which are determined by distances to new facility.Asymmetric rectilinear distance may be suitable to describe urban travel and so we adopt the distance in this paper.First we show the location problem can be reduced to a linear programming problem. The reduced problem is a special type and can be solved efficiently by Megiddo. Next we define an efficient point and clarify its property. That is, we focus candidates points of optimal location. Finally, we propose geometric solution procedure of the location problem.

View full abstract

This paper studies a knowledge finding method for large scale systems using a Fuzzy Classifier System (FCS). New payoffs and a new method for apportionment of credits for the FCS is proposed in this paper. The new method makes it possible to fully utilize the feature of the genetic algorithm, i.e. the effects of the crossover operator.Simulations to find rules avoiding mutual collision for a ship are done to show that the FCS can find fuzzy rules for complex tasks.

View full abstract

Generally, an expression of proposition for decision making has its later condition part which is inperative or conclusive such as "If S is A, then do B" or "If S is C, then T is D". But People often express a proposition by using auxiliary verbs of intention such as "If S is A, then you SHOULD do B" or "If S is C, then you MUST NOT do D" in thought.This paper proposes that we express a proposition for decision making by using auxiliary verbs of intention such as MUST, SHOULD, HAD BETTER and MAY and process it by an algorithm. The Algorithm makes a decision for actions expressed by verb by evaluating weight of proposition in order of strength depending on the auxiliary verb of intention. It is easy for designers to analyze process of decision making by this method.This method is applied to operation of autonomous rendezvous and docking for space vehicle. And the process of decision making for actions are shown by simulation.

View full abstract

When fuzzy causal relations and a fuzzy set of observed events are given, fuzzy abduction or fuzzy relational equations can be utilized to obtain a fuzzy set of causes of the given events. However, these techniques can not be applied, if the given problem includes causal relations in which events occur only when plural causes happen simultaneously. in this paper, we propose a method to apply fuzzy abduction to such problems. First, we express the causal relations with pliral causes as implications with antecedents of conjunctive composed propositions. Then, the composed propositions are replaced by temporal elemental propositions, and the conventional fuzzy abduction is applied to the problem with the temporal propositions. After truth values of the temporal propositions are obtained by fuzzy abduction, those of causes composing the composed propositions are derived. The problem to derive the truth values of causes from those of the composed propositions is called fuzzy Conjunctive Simultaneous Equations (FCSE). The paper discusses a way to solve FCSE, the strucure of solution space, necessarye and sufficient conditions of existence of solutions, and a method to get approximate solutions if the conditions are not satisfied. Finally, a numerical example is shown and its application to diagnostic problems is discussed.

View full abstract

The physical performance is defined as "The result associated with the motion inculde of the movement". To evaluate this physical performance is important in oder to improve current condition. Now, the physical performance is evaluated from two point of the view. One is from the physical fitness, which is measured from each physical elements(strength, muscle endurance, agility, balance, explosive strength, endurance, and so on). The other is evaluated from the skill, which is measured from skill test. But the total physical performance is generally evaluated by the supervisor or the coach depend on the experience which is based on these physical parameter. The reasons why such subjective evaluation have reliance for the evaluation of physical performance are that physical performance include the fuzziness derive from human and that human movements are too complex to describe effective means precisely. In this paper, to improve these problems, we propose the evaluation of the physical performance of human based on Fuzzy reasoning. By using Fuzzy reasoning, we can evaluate physical performance which include the fuzziness of human and also can have nonlinear evaluation depend on experience rule of expert with reliability.

View full abstract

Facial skin temperature can be measured in a non-contact manner using an infrared camera regardless of various types of living conditions. Moreover it varies with the autonomic nerve activity that accompanies arousal of the sensations, it serves as an effective physiological quantity for evaluating human sensations.In this report, a Lyapunov spectrum analysis evaluating orbital instability and unpredicatability of time series data and a correlation dimension analysis evaluating a kind of fractal dimension of time series data were applied to the nose skin temperature data in order to examine the effectiveness of estimating human sensations by these methods.For this purpose, the nose skin temperatures of 6 subjects at a relaxed state(closed eye) and stressed state (playing TV game) were measured in an experiment under a condition of 25℃, 50%RH, 1.0met(amounts of activity), 0.7,1.0clo.(amounts of clothing).Then the Lyapunov spectrum analysis and the correlation dimension analysis were applied.As a rsult, all the largest Lyapunov exponents and KS entropies o the relaxed state indicated significantly higher than that o the stressed condition. (P<0.01)Moreover, a correlation coefficient between these values and a sense of tension and excitement were significantly high.(P<0.05) On the contrary, the correlation dimension did not converge.Accordingly, it can be inferred that Lyapunov spectrum analysis for nose skin temperature is an effective method of evaluating human sensations.

View full abstract

It is very important problem to effectively construct a small-scale fuzzy neural network for multi input and output systems, because the number of intermediate unir functions exponentially grows as the number of input variables to the fuzzy reasoning increases, if some fuzzy sets are assigned to each input data. In this paper, four kinds of block hierarchical fuzzy-gaussian neural networks (FGNNs) are proposed for a control system of a mobile robot with two independent driving wheels by applying tow inputs and single output FGNN block, or single input and single output FGNN block. Such a block hierarchical FGNN mainly consists of three layers. That is, the first input layer consists of two FGNN blocks that independently generate torques for controlling the velocity and aximuth of the mobile robot. The second hidden layer determines their distributions to the final layer by using fixed connection weights.The final output layer also consists of two FGNN blocks that automatically determine the output scalers for the actual left-and right-wheel driving torques. The effectiveness of the proposed method is illustrated by performing the simulations of a circular path tracking control.

View full abstract

This paper presents a new algorithm for learning bang-bang controller which is described by fuzzy control rules. The then-part of the rules is weight vector for each control choice, such as "push right" and "push left".The fuzzy inference method is product for If-part, addition for Then-part, and selection for defuzzification. In the algorithm, the weight vector is tuned only according to the time until failure. By computer simulation, we apply the algorithm to the inverted pendulum problem for demonstrating the capability of learning the fuzzy control rules. We compare the learning ability of the algorithm with other algorithms, such as BOXES, AHC, CART, and IBRL3.

View full abstract

When tuning of fuzzy production rules which use fuzzy reasoning, the human experts handle the shape of membership functions conventionally, thus, there are several promising approaches for automatic membership function tuning. The mainproblem of these approaches is that, automatic tuning makes re-tuning the rules by a human expert impossible.In fuzzy production rules, the total matching degree of the condition part is calculated from the matching degree of each condition by aggregation operators. Most systems use "min" and "max" functions as aggregation operators. There is some research in the area where, the controller character is changed by using different T norms and T conorms, but this said nothing for tuning.In this paper, we consider an automatic operator tuning method for the parametric T-norms and T-conorms whose characteristics can be modified by parameters. By computer simulations, we found suitable operators that is intermediate between tipical operators, and T-conorms are not enough as combintion functions.

View full abstract

This paper considers the simulaneous analysis of classification and regression in the context of the fuzzy modeling. The fuzzy c-regression models (FCRM) by Hathaway and Bezdek is an extension of the classical fuzzy clustering to obtain a fuzzy partition of the given data and at the same time a number of regression models cprresponding to the data partition. Howerver, the direct application of this method to the fuzzy modeling dos not necessarily bring a good result. After explaining this fact using a numerical example, we try to revise the FCRM in two aspects : modification of the criterion and adaptation of the algorithm. An application to building of a fuzzy model related to the world population is presented to show how the proposed method works.

View full abstract

When the fuzzy the rule base is sparse, conventional fuzzy reasoning methods do not work well because of the lack of inference evidence. To tacke this problem. Koczy and Hirota proposed a fuzzy reasoning method called a linear interpolative method. In this paper, we analyze the Koczy and Hirota's reasoning mehtod and find out that the reasoning consequences by the method are sometimes abnormal fuzzy sets. Moreover, the reasoning conditions of the reasoning method are also discussed analytically.

View full abstract

The ultrasonic motor is a new type motor which is driven by ultrasonic vibration force of piezoelectric elements unlike other electromagnetic motors, and has many excellent performance and features. The ultrasonic motor, however, has speed ripple which arises from inherent natures caused by its driving principle and structure. Since this speed ripple is larger than that of other small motors, it is a significant problem in view of precise speed control.We paid our attention to the periodicity of speed ripple and reduced the speed ripple of ultrasonic motor by using repetitive controller with the fixed learning factor. However, it is difficult to reduce the speed ripple quickly and to carry out the precise speed control when the kearning factor is fixed. In this paper, we present a speed ripple reduction method to achieve the precise speed control by adjusting the learning factor using fuzzy reasoning. it is able to reduce the speed ripple to one tenth comparing with that without speed control.

View full abstract

The speed control ability of dc servomotors is affected by parameter variations and disturbance torque. Mathematical plant model is usually used for control, however, accurate parameter measurement is very difficult and parameters vary with operating conditions, for example, temperature or saturation effects. Consequently, motor control methods using the disturbance torque observer which compensates parameter variations and disturbance torque has been proposed, high-performance control of servomotor has been achieved. However, estimated disturbance torque by the disturbance torque observer has estimation error, therefore the control method taking account of estimation error is necessary.In this paper, the robust speed control method for a dc servomotor with a disturbance torque observer and a feedback controller is proposed. The disturbance torque observer is used for compensating parameter variations and disturbance torque. The influence of estimation error for disturbance torque is restrained by the feedback controller introducing fuzzy reasoning. The feedback again adaptive law in the feedback controller is constructed on the basis of the speed error information.The useful features and validity of proposed control method are confirmed by computer simulations.

View full abstract

Fuzzy reasoning methods are generally classiied into two approaches : direct approach and truth space approach. Several researches on the mathematical relationships between the direct approach and the truth space approach have been reported. There has been, however, no research which discusses the utility of the direct approach and the truth space approach.The authors have proposed three types of fuzzy neural networks (FNNs) celled Type I, II and III. The FNNs can identify the fuzzy rules and tune the membership functions of fuzzy reasoning atuomatically utilizing the learning capability of neural network. The TypeIII based on the truth space approach, especially, can obtain the linguistic fuzzy rules. But this type of FNN has some difficulties in understanding the fuzzy rules.This paper presents new types of FNNs called TypeIV and V based on the truth space approach. The fuzzy rules identified with the TypeIV can be comprehended more clearly than those with the TypeIII and the fuzzy rules of the TypeV can be expressed more flexibly than those of the TypeIII and IV. This paper also describes a method to label the fuzzy variables in the consequences with the linguistic truth values of the obtained fuzzy rules. The feasibility of the new FNNs are examined using simple numerical data. The results show that the truth space approach makes the fuzzy rules easy to understand.

View full abstract

Generally, a human being deals with various forms of data such as sound, image, numerical, and linguistic. In order to solve numerous problems which occur in our everyday lives, we usually use more than two forms of data. Unconsciously, a human being combines more than two forms of data and intefrates them and acquires knowledge from them. This process of a human being is called data fusion. In the human intelligent activities, language plays an important role. A human being uses language for thinking. Wittgenstein said that language is a tool for thinking. Language is important for human data processing. For the purpose of data fusion of heterogeneous forms of data, we regard language as a basic forms of data. Data processing based on language is called intelligent computing. In this paper, we consider an approach to realizing intelligent computing. Firthermore, we propose a framework of data fusion. In order to intefrate heterogeneous forms of data, the meanings of data must be considered. For the purpose of considering the meaning of data, we use systemic functional linguistic theory. According to this theory, the meanings of words are determined use of them. Using this theory, the meanings of heterogeneous forms of data are also determined use of them. Therefore, this theory is suitable for realizing the concept of the intelligent computing. In this study, we propose the framework of data fusion based on systemic functional linguistic theory.This framework is applied to the system which generates verbal commands a game called gate-ball. Acording to systemic functional linguistic theory, in the society where we live, there are laws, ideologies, morals which are called context of culture and they effect us in our sense of values. In the games, there are rules, strategies of their own, they also effect us in their restricted worlds. In other words, a game is a model of the society where we live. The system we propose in this study interprets rules, strategies, tactics of the gate-ball game and gives the instructions to the player suitable for the situations. In order to recognize the situations of the gate-ball game, not only numerical data but also image data and linguistic data must be intefrated. This system integrates them in the linguistic form. We give a certain situation of the gate-ball game to this system, describe the series of data proessing in this system, and consider the results of data processing.

View full abstract

Simplified fuzzy logic permits flexible representation of input-output relationships. In this study, sets of fuzzy rules based on simplified fuzzy logic were derived from observed data in the Fukushima area; meerorological data were input and rice yield data was output. These sets of rules outperformed a conventional multiple linear equation. Application of those rules to the ovserved meteorological data in 1993 predicts serious cold-weather damage of the year.

View full abstract