[in Japanese]

View full abstract

We introduce two kinds of approaches, a normative approach and a constructive approach, into design activities, and we characterize these two approaches in our mathematical formulation of design or education. And then, we show that there are significant relations between constructive approaches and emergent phenomena. Through analysis of constructive approaches and emergent phenomena, we give a way to represent results of emergent phenomena in our mathematical formulation of design or education. Moreover we analyze General Design Theory in our mathematical formulation of design, and we find a way to extend General Design Theory in order to deal with emergent phenomena in its extended frame.

View full abstract

In the hierarchical structure of an artificial system, the interaction and the attribute have the difference whether depend on the each layer of the hierarchy. Because the lower function is analyzed again as a higher function in the hierarchical structure of the function, one function is interpreted as two kinds (the interaction and the attribute).In this paper, the correspondence between "function as the interaction" and "function as the attribute" was mathematically analyzed by using the hierarchical structure in an artificial system. And, we showed mathematically that structural correspondence may not attach even if correspondence may intuitively attach. Therefore, it is thought that the attribute and the interaction are related by a complex structural mechanism. We want to insist that this phenomenon is emergence.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Simplified expressions are usually used in 2D drawings such as omissions and sectional views. Since the simplified expressions are not geometrical, they have been excluded in various problems such as 2D-3D conversion. In this paper, a method is proposed that 2D drawings including simplified expressions are automatically restored into exact drawings by using rules of the simplified expressions.

View full abstract

[in Japanese]

View full abstract

A support vector machines (SVM) is known as a pattern classifier with a high generalization ability and one of its advantages is that the generalization ability is theoretically guaranteed. However, many of the analyses are given in the framework of the PAC learning and the error-bounds are rather loose than the practical generalization error. In this paper, we present some studies on the average generalization error of SVMs, which is a more practical criterion for generalization ability.

View full abstract

[in Japanese]

View full abstract

Looking at the vast amount of literature, it seems that fundamental issues on nonlinear output regulation problem have been almost completely resolved. However, comparing with the results for linear systems, there still remains some important problems that need further discussions. Among them, in this paper, we consider the bounded-ness assumption for reference signals and robustness of output tracking property.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

We define the discrete-time state-reset system with the state-reset mechanism, which resets some elements of the state to zero when the elements exceeds the threshold.We show the state boundedness of the state-reset system including the property that the state has the invariant set.As one of examples of state-reset systems, We also consider the network-type decentralized control system and derive that the invariant set of the state estimation error exists.

View full abstract

[in Japanese]

View full abstract

A program package is developed for stability analysis of hybrid dynamical systems by computing Lyapunov functions based on SOS. Unlike existing tools, our package can treat vector fields that contain trigonometric polynomials. Lyapunov functions are automatically generated so that they satisfy conditions for each mode and each mode-transition.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

Since the year of 1977, the world of medical diagnosis has entered a new era with the invention of the modern magnetic resonance imaging (MRI) scanners. These scanners can produce high quality images of any part of the human body. MRI techniques are the method of choice in the diagnosis of tumors, and malformations of the brain because of the incredible ability of distinguishing between the different tissue types, and therefore providing valuable information about the problem at hand. In this thesis a method of segmenting an MRI data set of the brain and providing a 3D model of the white matter, the gray matter and the CSF is introduced. This method consists of three major steps: Skull Stripping, Tissue Classification and Volume rendering. The skull stripping method uses an anisotropic diffusion filter for noise removal, Marr-Helder edge detector to isolate anatomical boundries and a set of morphological functions to remove the skull from the data set. The extraction of different tissue types is then carried out according to the different signals each tissue type produces during the MRI process. This method is based on thresholding, and extracts all the pixels of the desired tissue type that fall within a specified interval. For the volume visualization of the results of the segmentation many procedures exist nowadays. This thesis uses the Visualization Toolkit, which is probably the most used one in the field of medical imaging. VTK provides the user with many libraries that could be used to suit the data type in question. For perfect visualization to be achieved, suitable functions had to be called and appropriate variables had to be set properly. Three accurate 3-D models of white matter, gray matter and cerebrospinal fluid were realized in this thesis according to the method described above.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

This paper presents an adaptive deadbeat control algorithm for the ABR traffic congestion control in ATM networks. We consider a control-based mathematical model and apply the tracking feedback deadbeat control design, online plant parameter identification and controller parameter modification algorithm to the problem. The practical control issues of output buffer saturation and initial conditions are considered in the control scheme. Simulations are implemented to verify the proposed control design. It is shown that the control algorithm satisfies the requirements of fairness to users, network efficiency, unknown time-varying delays and queue control. Moreover, the control algorithm is simple and a fast guaranteed convergence is obtained.

View full abstract

Recently, Buttazzo et al. address a feedback scheduling algorithm, which dynamically adjusts task periods based on the current load by applying a linear elastic task model. In their model, the amount of the resource allocated in each task is treated as the length of a spring with a constant elastic coefficient. Their approach guarantees the schedulability to avoid overload situations but does not take into account fairness of QoS levels. In this paper, we propose a nonlinear elastic task model which represents characteristics of the QoS of the tasks and perform a fair QoS management under the resource constraints for the purpose of avoiding overload situations. We also show the effectiveness of the proposed method by computer simulation.

View full abstract

Recently, an energy-aware resource management problem using dynamic voltage scaling technique has been attention to. In real-time systems, the execution of tasks with higher Quality of Service (QoS) requires more energy consumption as will as larger resources. On the other hand, to avoid deviation of execution performance among the current task set, fair-sharing of resource is needed. In this presentation, we formulate a trade-off between the maximization of achievable and fair QoS levels and the minimization of energy consumption as a nonlinear programming problem. It is shown that, using QoS fairness constraints, this problem can be solved efficiently as a single variable convex minimization problem.

View full abstract

Recently, many Quality of Service (QoS) adaptation control methods are studied to avoid overload conditions on soft real-time systems. We have proposed a QoS adaptation control method based on fairness of QoS levels. In this presentation, we implement this QoS adaptation controller on a real-time OS with a CPU time reservation function. We demonstrate a fair CPU allocation for a task set of many image trace applications with various QoS functions. It is shown that the proposed method can achieve fair QoS levels among applications dynamically.

View full abstract

[in Japanese]

View full abstract

[in Japanese]

View full abstract

In embedded systems such as mobile phones, home appliance and various control systems, software productivity is more required.This is partly because a wider variety of sensing information and more information processing capabilities are required in embedded systems.A community involving academia and industries needs and is established to discuss and solve not only the software issues but a whole range of embedded system technologies.

View full abstract

[in Japanese]

View full abstract

In this paper we treat a class of scheduling problems, the single-machine total tardiness problem with a precedence constraint. In this problem, a precedence constraint between two jobs is given a priori and the purpose is to obtain an optimal schedule satisfying this constraint. We construct a simple strict solution algorithm for this problem based on the extensions of the well-known decomposition theorem and the elimination rules for the ordinary single-machine total tardiness problem without a precedence constraint. It is shown by computational experiments that the algorithm can solve problem instances with up to about 45 jobs in acceptable times.

View full abstract

[in Japanese]

View full abstract

This study deals with proactive scheduling approach in which dummy jobs are inserted into a schedule in order to absorb impacts of uncertain events such as urgent jobs, machine breakdowns,and so on. This paper extends our previous work to flow shop scheduling problems with limited intermediate storages.We also propose an analytical model which determines a suitable amount of dummy jobs to be inserted into a predictive schedule.Some computational experiments demonstrate that the proposed proactive scheduling is effective by comparing it with predictable scheduling rules.

View full abstract

[in Japanese]

View full abstract