From applications point of view in real plants, we review the gain-scheduled control that is based on linear parameter-varying systems modeling. We also discuss two related topics in control problems for systems with time-delay and in LQG controls.
Dual control for uncertain systems considers the dual roles of control inputs for control and online estimation, adding probing signals to control inputs. The control inputs can be precisely optimized by using stochastic dynamic programming, considering all the uncertainties in the control system. This is called “optimal adaptive control”, but this requires a vast online computation. Several years ago, some researchers proposed “dual predictive control”, which approximately optimizes control inputs with nonlinear optimization, requiring a reasonable amount of computation. Recently, the control has been improved to consider the uncertainties in future inputs and output references. In dual control, probing signals behave chaotic, and they should not be too large. In this sense, dual predictive control can be considered to be along, or to move towards, “the edge of chaos”, in terms of the complexity science.
In this paper, the passivity-based design principle of adaptive controller for nonlinear systems is described. The relation between the passivity and the stability of nonlinear system is introduced. Using the relationship, two kinds of design approaches, cancellation and domination control, are discussed.
In this paper, we consider a dynamical model of computer networks and derive a synthesis method for congestion control. First, we show a model of TCP/AQM (Transmission Control Protocol/Active Queue Management) as a dynamical model of computer networks. The dynamical model of TCP/AQM networks consists of models of TCP window size, queue length and AQM mechanisms. Second, we propose to describe the dynamical model of TCP/AQM networks as linear systems with self-scheduling parameters, which also depend on information delay. Here we focus on the constraints on the maximum queue length and TCP window-size, which are the network resources in TCP/AQM networks. We derive TCP/AQM networks as the LPV system (linear parameter varying system) with information delay and self-scheduling parameter. We design a memoryless state feedback controller of the LPV system based on a gain-scheduling method. Finally, the effectiveness of the proposed method is evaluated by using MATLAB and the well-known ns-2 (Network Simulator Ver.2) simulator.
It is important to find a method using the dielectric properties of the material to determine moisture content independent of the density in developing microwave moisture meter. Moisture measurement based on the dielectric properties of the material can be applied in various microwave techniques. In this paper, a new method using dielectric constants only for density-independent moisture determination is proposed. The new method was used to measure moisture content of green tea, and the results are presented. It was shown that the Root Mean Square Error (RMSE) is 2.4% in the moisture measurement for the samples in 8-49% moisture range and 0.14-0.69 g/cm3 density range. The comparison of the proposed method with the former method on green tea was also presented. The results show that the accuracy of moisture determination by the proposed method is better than the former method.
This paper proposes a method to improve the contrast of a color image using a genetic algorithm. The accumulated edge strength in the R, G and B channel images that compose an output color image is used for evaluating an individual fitness. The difference of color hues between the original color image and the output color image is also used for evaluating the fitness. The genetic algorithm works to generate an output color image that has large edge strength in the R, G and B channel images and has less difference of the accumulated hues between the original image and the output image. The experimental results show that the output color image by the proposed method had good contrast visually compared with the results by the liner brightness transform method.
In signal processing and image processing, many filters have been studied for smoothing images corrupted by noises. In particular, some nonlinear filters have been investigated. In nonlinear filtering, a neural filter was proposed for reducing Gaussian noise. However, a neural filter was not applied to images with impulse noise. In this paper, we propose a filter for removing impulse noise, Gaussian noise and mixed impulse and Gaussian noises using median filters and neural filters. Carrying out the simulation, we illustrate the performance of the proposed filter.
In this paper, a new effective algorithm for extracting spectacles from facial images and generating of the face without spectacles is presented. In the biometric person authentication through facial images, many faces with spectacles must be taken into account. However, the verification of the two images between one face with spectacles (input image, for instance) and another face without spectacles (registered image, for instance), which are taken from the same person, are not considered in the conventional system and have not been investigated thus far. In the practical case, however, these conditions are frequently considered. To expect a high matching ratio and stability for distinguishing a person from others, an effective method of isolating spectacles from facial images with spectacles must be developed. From this point of view, an effective method to isolate spectacles from facial images and the results of the experiment that occurred are presented. By using many types of spectacles, the efficacy of the proposed method was examined. As a result, 80.7% of extraction accuracy was obtained.
Lineaments are important features showing subsurface elements or structural weakness and are in general extracted by visual interpretation of experts. Lineament maps given by the visual interpretation should have a gap because of a poor contrast of the segments in data. To generate lineaments, which are collinear and broken into a series of the segments that obtained from multispectral data of Landsat Thematic Mapper (TM), this paper proposes an automated lineament detection using land cover information in mixel (mixed pixel). Many lineaments can be extracted from a drainage system for topography. Therefore, elements of water system that estimated with the fuzzy reasoning was applied to synthetic edge data from first- to fourth- band data. In addition, the edge in the region of town and paddy field, which are no relation to lineaments, was excluded. And then, we generated segments from the resulting edge data considering the feature in each band data. As a result of experience, it is clear that the proposed approach using land cover information in TM data can permit reliable extraction of segments influenced by geological structures.
This paper deals with H∞ control attenuating initial-state uncertainties, and its application to a magnetic suspension system. H∞ control problem, which treats as a mixed attenuation of disturbance and initial-state uncertainty for linear time-invariant systems in the infinite-horizon case, is examined. The mixed attenuation supplies H∞ controls with good transients and assures H∞ controls of robustness against initial-state uncertainty. We apply this method to a magnetic suspension system, and evaluate attenuation property of the proposed disturbance and initial-state uncertainty via simulations and experiments.
The problem under consideration centers on building a three-dimensional description of the unprepared environment of an autonomous mobile robot. In an image sequence, tracking is to be performed after image rectification. This intermediate process minimizes the token relative displacements between two frames and simplifies the tracking phase, because it reduces the disparity between two relative tokens and thus simplifies the matching process. In this paper we present a new hybrid approach to range estimation that combine inertial and visual based technologies; this allows us to calculate the image-space distance between the robotic head and 3D tokens. Two frames from the image sequence obtained from passive-target-tracking system, moving CCD video camera, will represent a set of data with the output of the inertial-tracking system, that report the relative changes of orientations and accelerations between the two frames. By integrating these data in our algorithm the image-space distances of different 3D points was estimated theoretically and experimentally.
The aged and disabled constitute a growing percentage of the world population, and a variety of systems to assist them are coming into very high demand. The purpose of our study is to develop a power wheelchair that gives the aged and disabled the same degree of mobility that healty poeple enjoy, enabling users to rejoin society fully and heartily. To accomplish this, we adopt a holonomic omnidirectional mechanism that provide 3 DOF mobility, the same as healthy people have. In addition, we introduce an intuitive interface and automatic control functions to the power wheelchair. With these technologies, a power wheelchair can provide flexible and intricate motion through simple commands.
It is important to measure the linearity of a welded part to evaluate the quality of manual welding. This paper proposes a method to measure the linearity of a welded part that has an uneven surface condition by image processing. The initial template image that includes a part of a welded area is set at the left edge in the image. Next, the template image matching is repeated to the right edge continuously. In the template matching process, the template image is renewed to the searched image pattern at every matching location. The vertical differential position of the welded part is measured in the continuous template matching process and their standard deviation is calculated as the linearity. The experimental results show that the curve obtained by the proposed method expresses the global form of the welded part appropriately. The linearity measured by the proposed method correlated closely with the linearity measured by the visual sense.
In this paper, a Hebbian learning rule restraining “catastrophic forgetting” is proposed on pulse neural network (PNN) with leaky integrate-and-fire neurons. The strong point of this learning rule is that a learning of new pattern does not destroy past ones, and that an efficient use of synapses is enabled. First, in order to consider the function of the learning rule, a fundamental experiment is made. Next, to compare the performance between the proposed learning rule and conventional ones on the application, simulation experiments are examined using autonomous behavior robots which are forced to learn concurrently two different environments. The results of the experiments show that the proposed learning rule clearly restrains “catastrophic forgetting” and enables working of more efficient than conventional PNN learning.
One of the most important issues for power suppliers in the deregulated electric industry is how to bid into the electricity auction market to satisfy their profit-maximizing goals. Based on the Q-Learning algorithm, this paper presents a novel supplier bidding strategy to maximize supplier’s profit in the long run. In this approach, the supplier bidding strategy is viewed as a kind of stochastic optimal control problem and each supplier can learn from experience. A competitive day-ahead electricity auction market with hourly bids is assumed here, where no supplier possesses the market power. The dynamics and the incomplete information of the market are considered. The impacts of suppliers’ strategic bidding on the market price are analyzed under uniform pricing rule and discriminatory pricing rule. Agent-based simulations are presented. The simulation results show the feasibility of the proposed bidding strategy.
In June 2001, the International Agency for Research on Cancer (IARC) classified ELF magnetic fields as a possible human carcinogen based on the results of two recent pooled analyses by Ahlbom et al. and Greenland et al. on epidemiological papers for childhood leukemia. Examining the data displayed by them, this author succeeded to read the other informations than theirs: in the paper of Ahlbom et al., for instance, 9 cases were prepared for verifying the hypothesis that magnetic fields are associated with childhood leukemia, and this author considers that the general conclusion derived by inductive inference is that the hypothesis is not supported, because 2 cases out of 9 support the hypothesis and 7 cases do not support the hypothesis, although the hypothesis has been believed by many people for about twenty years; furthermore, the author does not consider that the IARC’s evaluation that risk doubles in excess of 3 or 4 mG is based on the general conclusion, because it comes from pooling two kinds of data, one of which is of 7 cases bringing the ‘general conclusion’, and the other of the remained 2 cases being the ‘particular facts’.
The application of chaos to global optimization methods are, 1) maps with respect to inner variables derived by discretizing gradient method models with Euler’s method are unstabilized by setting their sampling time large, 2) chaotic trajectories of the optimizer’s variables confined in the bounded searching domain are generated by nonlinear transformations of the unstabilized inner variables, and 3) the chaotic annealing method is available which conversely stabilize their dynamics by gradually decreasing the sampling time of them. However, the efficiency of the above-mentioned method called the chaotic global optimization was only reported for optimization problems constrained by upper and lower bounds. In this paper, to the contrary, we attempt to apply the method to optimization problems with normalized equality and non-negativity constraints. First, based on the replicator model which is regarded as the gradient projection method with a variable metric, two types of chaotic maps on a simplex are presented. The one is a discretized steepest gradient model with respect to inner state variables to which the replicator model is equivalently transformed, and the other is a discretized replicator model for an unconstrained problem with respect to inner state variables obtained by variable transformation. Secondly, the bifurcation with respect to the sampling time of Euler’s method is shown and the feasibility of the trajectories on a simplex for normalized equality and non-negativity constraints is certified. Lastly, the chaotic global optimization methods with the annealing procedure are demonstrated in numerical simulations for a few constrained optimization problems.
This paper deals with spatio-temporal indexing method for moving objects. In our research, we propose XAT (eXtended Adaptive Tree) structure, consisting of spatial trees and temporal trees, for fast search for spatio-temporal data. The searching process in XAT structure is divided into two steps. The first step roughly narrows down the potential solutions (moving objects) according to the given searching range. The last step fixes the real solution by checking the object’s moving track. We compare XAT structure and 3D structure, one of the conventional methods, by computer simulation. The result shows that XAT structure works faster than 3D structure when there is difference between the spatial search range and temporal search range and the objects’ moving areas are small.
Investors or fund-managers face with optimization of portfolio selection, which means that determine the kind and the quantity of investment among several brands. We have developed a method to obtain optimal stock’s portfolio more rapidly from twice to three times than conventional method with efficient universal optimization. The method is characterized by quadratic matrix of utility function and constrained matrices divided into several sub-matrices by focusing on structure of these matrices.
We have already proposed a methodology for static inverse optimization to interpret real data from a viewpoint of optimization. In this paper we propose a method for efficiently generating constraints by divide-and-conquer to interpret large-scale data by static inverse optimization. It radically decreases computational cost of generating constraints by deleting non-Pareto optimal data from given data. To evaluate the effectiveness of the proposed method, simulation experiments using 3-D artifical data are carried out. As an application to real data, criterion functions underlying decision making of about 5, 000 tenants living along Yamanote line and Soubu-Chuo line in Tokyo are estimated, providing interpretation of rented housing data from a viewpoint of optimization.
We propose a new image enlargement method employing a codebook-based fuzzy interpolation technique in order to realize simultaneous smoothing and sharpening in image enlargement. In the proposed method, codebooks are generated by using self-organizing maps (SOM). The codes of the generated codebooks are used to determine parameters of Gaussian membership functions in the fuzzy IF-THEN rules. Then, enlargement is achieved by a fuzzy inference. Experimental results of the proposed method show superior performances than other typical image enlargement methods. Calculation time is reduced drastically in comparison with the previous codebook-based method which has been proposed by the authors.