Abstract:
PROBLEM TO BE SOLVED: To automatically discriminate whether to apply sliding mode control in a plant from minimum required measured experimental data.SOLUTION: A system first measures a plant in an open control manner and uses the measured data to describe a state equation of the plant in accordance with an existing system identification technique and an order determination method. A switching hyperplane of sliding mode control is designed. Next, regarding a difference between output in the case where only control input of a linear model restricted by the hyperplane is used, and output in the case where control input of a sliding mode is used, a higher-order statistic of a third-order moment and a fourth-order moment are calculated. In the case where a value of the third-order moment is greater than a predetermined threshold value or a value of the fourth-order moment is greater than the predetermined threshold value, a controller is configured as a sum of a linear control input term and a non-linear control input term. In the case where it is determined that both the value of the third-order moment and the value of the fourth-order moment are smaller than the predetermined threshold value, the system uses only the linear control input term to configure the controller.
Abstract:
PROBLEM TO BE SOLVED: To configure a semi-optimum robust PID controller when input and output are observed even in an instable system lacking the advance information of a plant.SOLUTION: A function is given to perform narrowing from the coefficient space of a PID controller excluding a combination disconfirmed by input/output data, move a candidate group while sharing the information of the evaluation function history of each candidate in the coefficient candidate group and an evolution history as a whole, and automatically design in a non-disconfirming manner a coefficient which achieves robust stabilization in the entire system of the plant and the controller. While storing/sharing the information of a PID gain which is the minimum in the evaluation function history of each candidate and a PID gain which is the minimum evaluation value in the candidate group whole history, the whole candidate group is moved aiming at a better center by PSO, automatic design is performed by executing a candidate peripheral search and disconfirmed candidate elimination from it, and non-disconfirming control is executed even to the peripheral search candidate of respective particles of the candidate group by the PSO.
Abstract:
PROBLEM TO BE SOLVED: To automatically distinguish a simplifiable part of system representation by a quantitative reference, in a modeling tool for a design. SOLUTION: An expression is extracted from a block of SysML, or a block diagram of Simulink (R) is converted into a differential equation, and a coefficient thereof is expanded in a series. First, the differential equation is normally solved by a CAS, and preferably, a solution thereof is also expanded in a series. Next, the expression including the differential equation is solved by the CAS as an interval coefficient by interval analysis technique. Coefficient vectors of the stored solution obtained by performing the normal solution and the solution of a result of the interval analysis are compared. As a result, when an inter-vector distance is a fixed value or below, a target coefficient is set to zero, and the expression including the differential equation is solved by the CAS. The coefficient vectors of the stored solution obtained by performing the normal solution and the solution of the result are newly compared. As a result, when the inter-vector distance is the fixed value or below, a term of the target coefficient is dropped from the differential equation to simplify the differential equation. COPYRIGHT: (C)2011,JPO&INPIT
Abstract:
PROBLEM TO BE SOLVED: To stably separate and extract an original signal from more than one signals by a few calculation steps when the signals in a mixed state are observed. SOLUTION: A function having a monotonous increase characteristic such as an index function is introduced as an evaluating function and the signal is separated by applying an adaptive algorism for minimizing the evaluating function for a signal separation matrix. That is, a signal processor separates the original signal from an observation signal x(t) and outputs it. A non-linear function 21 works with respect to the inputted observation signal x(t) and an estimated separation matrix W(t-1) which is estimated in a preceding cycle and a differential signal e(t) is calculated (22) based on a value y(t) generated by the non-linear function 21, the estimated separation matrix W(t) and the observation signal x(t) at this time. Differential evaluation is weighted by the evaluating function having the monotonous increase characteristic based on the calculated differential signal e(t) and, then, the separation matrix W(t) at that time is updated (23).
Abstract:
PROBLEM TO BE SOLVED: To obtain clear enlarged or reduced images without damaging the rough shape of a font, etc., with respect to a graphic picture including many step edges such as fine lines handled by an information display such as an LCD panel and a projector. SOLUTION: This image processing apparatus for generating transformed image data by performing resolution conversion that enlarges/reduces original image data into a multiple of a rational number or an integer is provided with an original image data inputting part 11 for inputting the original image data, an algorithmic nonlinear processing part 12 for generating first transformed image data maintaining a connection relation to line width held by the inputted original image data even after the resolution conversion, a multiage processing part 13 for generating second transformed image data subjected to resolution conversion in the direction of maintaining the rough shape of the line width with respect to the inputted original image, and a synthesizing part 15 for combining the generated first transformed image data and the generated second transformed image data on the basis of the weighting by a weight deciding part 14.
Abstract:
PROBLEM TO BE SOLVED: To make an input image into halftone without damaging the feature character of an original image and to unnecessitate much arithmetic quantity increase by changing an error diffusion coefficient corresponding to each pixel in the image for each pixel. SOLUTION: A block 230 is provided with an error diffusion filter means, a block 250 is provided with a feature extracting means, a block 260 is provided with a threshold operating means, a block 270 is provided with an adaptive algorithm means and a block 280 is provided with a weight coefficient control means. Then, both feature extracting information and the quantization error information of input/output are provided, while including them with the differential norm of input/output information as a standard and by trading off both the information, a diffusion coefficient is controlled. At controlling of the error diffusion coefficient, a result (v) (n1 and n2) of feature extraction, input information (x) (n1 and n2) and output information (y) (n1 and n2) are utilized. A value λ(n1 and n2) adjusted with respect to each pixel is multiplied to the output, for which threshold processing is applied to a diffusion filter. Thus, the error diffusion coefficient corresponding to each pixel in the image is caused to change for each pixel.
Abstract:
PROBLEM TO BE SOLVED: To provide a technique for automatically configuring a controller with respect to each candidate for a scheduling parameter vector.SOLUTION: An LPV model of a plant is described, the number of scheduling parameters is given, and a maximum value and a minimum value found from a high-order request of restriction conditions are given to each element of a scheduling parameter vector comprising a plurality of elements. Each candidate has a ν gap calculated among Min, Max, and an intermediate value between them respectively, respective candidates are determined as scheduling parameter candidates in the decreasing order of differences between ν gaps. The number of parameters given first is selected and converted into an LMI, the controller is found based on an Hbasis for all combination end points associated with the maximum value and minimum value of the selected scheduling parameter, and a controller for a generalized plant is constituted by bilinear interpolation weighted with the ν gap calculated during the parameter selection.
Abstract:
PROBLEM TO BE SOLVED: To cope with a problem in which a method for predicting a risk factor latent in a project is required. SOLUTION: A computer system is provided for performing risk prediction in a project. The computer system includes: a collection part which collects answer data to a plurality of questions related to each project; a degree-of-risk assignment part which assigns, based on answer data to each question for each key item, a degree of risk to the key item; a learning part which learns relations between a plurality of answer data and a plurality of degrees of risk in a neural network; and a narrowing part which gives answer data including intentional answer data to the learnt neural network to determine a degree of risk, and narrows a question corresponding to the intentional answer data with high degree of risk as an important question; and a detection part which selects a project including answer data which increases the degree of risk of answer data to the important question, and performs risk prediction of a predetermined period with respect to the selected project. COPYRIGHT: (C)2010,JPO&INPIT
Abstract:
PROBLEM TO BE SOLVED: To optimize a size of a buffer pool for storing contents of a data base temporarily. SOLUTION: The data base server is for optimizing the size of the buffer pool and comprises: a division ratio setting part for setting division ratio by which a storage area of a predetermined size is divided into a plurality of buffer pools to be selectively used in accordance with the kind of transaction or the kind of data; a response time measuring part for measuring response time of the transaction under the state that the division ratio is set; a response time computation formula generating part for calculating a parameter of a response time formula for computation to be a formula for calculating the response time on the basis of the respective sizes of the plurality of buffer pools on the basis of the set division ratio and the measured response time; and a division ratio calculating part for calculating the division ratio to minimize the response time on the basis of the response time calculation formula of computation generating part. The division ration setting part provides the data base server to set the division ratio calculated by the division ratio calculating part. COPYRIGHT: (C)2006,JPO&NCIPI
Abstract:
PROBLEM TO BE SOLVED: To provide a technique of improving estimation accuracy in an unscented Kalman filter on the basis of nonuniform update timing which considers plant characteristics.SOLUTION: The system achieves nonuniform updates by input to a Kalman filter through an all-pass filter in which a relationship between a numerator polynomial and a denominator polynomial is described as a relationship between mirror image polynomials. Thereby, the input reduces the amount of computation that is different from re-sampling using a polyphase filter. Next, since an estimated quantity of state obtained by the Kalman filter has a nonuniform phase interval if no processing is performed thereon, it is filtered through inverse transformation of the nonuniformity.