In this paper, the flexible widget layout (FWL) problem is formulated as a fuzzy constraint satisfaction problem (FCSP). Widget layout is one of the most important challenges for automatic generation of graphical user interfaces (GUIs). In the field of model-based user interface design, the widget layout is more complicated because the process of selecting widgets is needed. We named this automatic GUI generation accompanied by the widget selection, FWL. The FWL requires both (1) deciding which widget types and their alignments are used and (2) completing the layout in a certain time especially when the system generates them in run time. Our system automatically selects appropriate widgets and lays them out in certain rectangles. We formulate the desirability of the selection straightforward as fuzzy constraints; therefore, we can utilize existing general technique of FCSP for the FWL. We divide the layout process into three phases, and using an existing FCSP algorithm, we realize the layout in a short time enough not to keep users of the generated GUI waiting.

View full abstract

Recently, the rapid progress of computers introduces evolutionary computations to next step, which is the demand for the variety of Pareto solutions in multi-objective optimization problems. We can acquire a large amount of Pareto solutions in a short time. However, it is difficult to grasp the acquired Pareto solutions effectivelly, because the Pareto solutions have multi-dimensional fitness values. This paper tries to support a user to select satisfying solutions with visualization method. This paper applies a visualization method to acquired Pareto solutions which have multi-objective fitness values. It enables us to grasp the distributed structure of Pareto solutions and clarify the relationship among multi-objective fitness values. This paper shows that the visualization result enables us to interpret the characteristics of Pareto solutions through experimental result of Nurse Scheduling Problem (NSP) which is one of the multi-objective optimization problems.

View full abstract

Immune Algorithm (IA) is an optimization algorithm which searches for multiple optimal and semi-optimal solutions by antibody reproduction and self suppression. This paper focuses on Bin Packing problem (BP), one of typical optimization problems, in which the effectiveness of IA has not been verified yet. This paper proposes IA to solve BP, discusses parameter configuration ways, and compares the search performance between existing genetic algorithm (GA) and IA. Experimental results have shown that IA could find optimal solutions even in the problems GA cannot solve.

View full abstract

Recently, a novel method for solving combinatorial optimization problems is proposed. Called a virtual magnetic diminuendo (VMD) method, it is originally inspired by an analogy between the Hopfield network and the spin glass, and provided just controlling a threshold corresponding to a magnetic parameter. According to the preliminary study, it is confirmed that the rate of reaching an optimal solution is improved through introducing the virtual magnetic parameter. But, at the same time, another comment that changing its polarity from negative to positive is essential part of the VMD method is also pointed out. As described above, a detailed mechanism of the VMD method has not been clear yet. Then, in order to solve this severe problem, two kinds of tasks such as an N queens problem and a number place problem are adopted, and we examine their performance under the some different scheduling functions of the virtual magnetic parameter. As a result, it is found experimentally that all trials with changing polarity show better performance than that without changing polarity.

View full abstract

In this paper, we propose an evolution-based method for controlling robots using evolutive binary decision diagrams (BDDs). BDDs are a way of representing a logic function using a directed graph that consists of binary switches having one input and two outputs, and are mainly used to design LSIs. Moriwaki et al. proposed n-output BDDs (n-BDDs) that can evolve BDDs, and they applied the n-BDDs to a food chain simulation of evolutional agents. However, in the evolution method, variable ordering of n-BDDs is fixed, so it cannot be applied to complex problems. In this paper, we introduce a new genetic manipulation to n-BDDs as a way of avoiding this problem of variable ordering, and we apply the new n-BDDs to obtain motion control of robots: LEGO Mindstorms and the humanoid robot HOAP-1. We also demonstrate that our method is capable of providing good performance in computer simulations.

View full abstract

Japan has frequent natural disasters, including typhoons and earthquakes. Nevertheless, building or civil structures are becoming taller and longer in recent years because of advances in civil engineering technology. Such structures are likely to suffer from vibrational problems caused by external loads. Therefore, it is extremely important to control structural vibration to improve the serviceability and safety of such structures. To develop a practical structural control system, it is necessary to consider their special features such as complexity, uncertainly, and size. To consider these features, fuzzy theory has been applied to structural control systems. These actuator systems are generally set to the top floor of the structure to reduce the vibration relating to the 1^st mode. However, in case of controlling the vibration of a multi-degree-of-freedom structure it is extremely difficult to reduce effectively response acceleration of the middle or low stories by only setting up an actuator system to the top floor. Also, although setting up actuator systems to each story can reduce effectively response, such idea can not be realized because of economical problems. To overcome these technical hurdles, this study attempts to locate the actuators dispersively at some floors by considering the costs of setting up actuator systems and their performances. In this study, Particle Swarm Optimization is introduced to optimize simultaneously the location and performances of actuator systems, under the cost constraints. The effectiveness and applicability of the proposed method are demonstrated through the numerical simulations.

View full abstract

The estimation of accidents at the intersections would be required in the traffic safety. Many factors have been specified to describe the mechanism of traffic accidents. At the same time, the estimation with high accuracy would be proposed considering the impact of road maintenance works. The model with neural fuzzy model is introduced to develop the estimation model. The genetic algorithm is applied involving the neural fuzzy model to determine the parameters in the model. It is concluded that time and spatial distribution of traffic accidents are estimated accurately with the proposed procedure.

View full abstract

In this paper, we consider the optimization of hierarchical systems where two decision makers with different priority exist in stochastic environments. To be more specific, formulating them as two-level linear programming problems where right-hand constants in constraints are random variables, we reformulate them as two-level simple recourse problems. For these reformulated problems, we attempt to apply interactive fuzzy programming in order to derive satisfactory solutions for the decision maker at the upper level in consideration of balance between the satisfactory level to the decision maker at the upper level and that to the decision maker at the lower level.

View full abstract

This paper proposes a perceptual system based on the concept of perceiving-acting cycle to realize the situation-dependent perception for a partner robot. Recently, various sensors enable an autonomous robot to get much information on the environment easily. However, the robot must extract the information suitable for the situation from such much information to take a flexible action like a human. In this study, we emphasize the importance of human vision in order that the robot flexibility realizes perception and action. Especially we focus on the perceptual system based on perceiving-acting cycle discussed in ecological psychology. First, we propose a retinal model based on human retinal structure, and a method for the information extraction using a spiking neural network based on perceiving-acting cycle. Next, we apply the proposed method of a partner robot with a laser range finder for a human tracking task. As experimental results, we show that the proposed method can directly perceive the necessary information based on the retinal model. Furthermore we show that the spiking neural network can realize the prediction-based perception and can improve the noise tolerance in the perception.

View full abstract

This paper describes a nursing fuzzy expert system for preventing pressure ulcer development in post-operative patients. First, knowledge of expert nurse for pressure ulcer development is derived. Second, a fuzzy expert system is constructed by a simple calculation of fuzzy membership functions expressing the knowledge. The expert system produces a fuzzy degree by averaging the calculated values obtained by these membership functions. This system performance is evaluated by leave-one-out cross validation on 10 patients. Then, the nursing fuzzy expert system can determine the pressure ulcer development with better precision, compared with the past work with statistical analysis on 360 patients.

View full abstract

The purpose of this study is to assess the improvement of production time in the field of cell production systems in a flat organization without a leader. In this study, a performance-maintenance (PM) action discovery table, based on the PM theory, was used for appointing an informal leader. In addition, a worker-machine evaluation table, based on structured heuristic evaluation methods, was used for evaluating the usability of machines in the cell. The main results were that (1) production time decreased by about 7.5% and (2) knowledge management was achieved by sharing individual knowledge and information from the PM-action discovery table and the worker-machine evaluation table. These results show the effectiveness of knowledge management provided by an informal leader on the improvement in group achievements.

View full abstract