The transition of intelligent systems from laboratories to real world in the last decade has created a need for humanization of intelligent systems. In this paper we outline human-centered framework for development of intelligent systems. A distributed multi-layered multi-agent architecture is one of the four components of the framework. We describe in this paper that how the multi-layered architecture facilitates human-centered modelling of intelligent systems from multiple perspectives. The framework and the distributed multi-layered multi-agent ontology has been applied in a range areas including alarm processing, web mining, image processing, sales recruitment and medical diagnosis.
An approach towards Emotional Intelligence implemented in Computer Mediated environments that improves the communication between humans via computer-aided devices as well between humans and machines was developed. Our greatest efforts were paid to how to extract, transfer and express emotion. Disregarding the semantic components, we examined different modes of communication: audio, visual and motional. Our approach is based on physical parameters, human based emotion feature extraction and model based decision method. For each mode the emotion relevant parameters were defined, extracted and their number reduced. Emotion samples were "translated" and "coded" into meaningful feature parameters. The relation between the physical variables and the psychological attributes was examined, and then the emotional groups based on parametric models were defined. Application examples are presented and the future direction of the affective human centred computing is discussed.
In this paper, a diagnosis system, using fuzzy logic, for rotor bar defects in the squirrel cage induction motor is proposed. In the iron and steel industry, induction motors are widely used for various facilities. Recently induction motors have been used in rolling mills. Accordingly the requirement for stabilization in their operations has become very stringent. Various methods are used to diagnose the rotor bar defects of the induction motor, such as the frequency analysis method, the neural network method and so on. However it is difficult to quantitatively determine the degree of deterioration by conventional diagnosis methods. To overcome the difficulty, a new identification method using fuzzy logic is proposed for the diagnosis of rotors in the squirrel cage induction motor. Due to a lot of non-linear elements in the induction motor model, conventional identification methods, such as the least squares method, are not applied to the identification of the induction motor parameters. Our proposed method utilizes the simulator based on the mathematical model and the actual data. The rotor bars resistance in the simulator is modified reflecting the difference between the calculated current and the measured current. After iteration of these modification steps, the rotor bars resistance in the simulator converges to its real value. Using our proposed method, the trend of the rotor deterioration can be managed quantitatively, enabling the appropriate condition-based maintenance of the induction motors.
Conjoint analysis is a method for deriving the part worth value of each factor from the total evaluations. In conjoint analysis, it is generally assumed that the value function is pre-defined as the additive and transitive model. However, there exist many cases in which the pre-defined value function is not appropriate to real situations. In this paper, we propose conjoint analysis based on rough sets approximated by Dominance relations. In our method, preference structure in decision making problems can be represented by IF-Then rules and better combinations of factors than the given ones in the given information system can be found by partial order relations.
This paper proposes a new prediction model of typhoon position introducing a hierarchical combination of decomposed fuzzy model, peripheral wind fields around typhoon, and categorization. This model has the feature that a multiple inputs-single output system is expressed by the hierarchical combination of multiple single input-single output fuzzy reasoning units. This produces the merits of less number of fuzzy rules and simplicity of rule construction. We describe a model construction process, applying principal component analysis and multi-regression analysis skillfully to a bulky data of wind field data and time series data of typhoon track. Finally, the prediction performance of the proposed model is compared with the PC method and Typhoon model, which are developed by Japan Meteorological Agency. It is shown that the proposed model is superior to the PC model but inferior to Typhoon model.
From a viewpoint of real engineering applications, a new and important concept of neighborhood is proposed for the solution of COP like VRP or TSP, which has been re-formulated with fuzzy set theory. Since the known cost information between the elements of COP is normalized into the real number [0,1], A concept of neighborhood degree is also proposed to measure the scale from the nearest (smallest cost) to the farthest (biggest cost) with fuzzy evaluation. A total neighborhood measure is proposed for estimating the quality of the solution in dynamic exploring process, and a partial neighborhood measure is also proposed, where the inferior portion of the worst solution can be detected and improved in the next operation. The properties and main role of the whole neighborhood measure are shown by the TSP (Traveling Salesman Problem) experiments, which the various tour data (benchmark, random and fractal type) are used and the scales are adjustable at 30-5000. Another experiment with dispatch and delivery problem for real trucks is also performed, where the poor tour (for vehicles) is detectable by the whole neighborhood measure and the poor trip (sub-tour for delivery job) can be caught through partial neighborhood measure. The algorithm using above strategy can avoid unnecessary parameter setting and press the useless and duplicate operation down. It is confirmed that the whole computational process is 15-30% faster than usual evolutionary method.
A method is proposed and discussed to evaluate credibility of a combined result of plural observations with uncertainty and contradiction for the cases of positive dependence, negative dependence and independence among the observations. In this paper credibility of an observation is described by 4-tuple grades, I.e., grades of truth, contradiction, uncertainty and false in the observation. These grades are determined in the same manner as membership functions are determined based on a concept of ground set. Then the credibility of plural observations is evaluated by the combination of fuzzy set products, where the fuzzy sets are characterized by the above mentioned grades, and the set products are determined using general probabilistic set operations. Credibilities for a pair of observations and repeated plural observations are derived and their properties are discussed.
The Emotion Generating Calculations (EGCs) can calculate human's emotions based on personal favorite values. The favorite values indicate the degree of likes/dislikes for some predefined objects. In this paper, we apply the "limitation of fuzzy truth value" method for evaluating the impression of noun clause. We consider that the relationship between the "truth value of fuzzy proposition" and the "linguistic truth value" is the same as that between a modified word and the modifier clause. An object has two types of truth values for like and dislike. The validity of this method is examined for some examples of noun modifier with verb in Japanese fairy tale "KACHIKACHIYAMA."