Fuzzy theory has been broadly used to represent information with uncertainty, because it is useful to deal with classes whose boundaries are not sharply defined. Fuzzy concept has remarkably expanded its application fields through fuzzy control, expert system and etc., because it can handle, in particular, fuzziness which arises when knowledge and know-how are represented on a computer. Nowadays, fuzzy theory has been going toward the maturing stage after passing through the beginning and growth stages in the point of view of application. To make its maturing stage run long, it is necessary to develop and to find various application fields of for fuzzy theory. In this talk, we discuss about some ideas for applying, widely and properly, fuzzy theory to marine industry and introduce applications and new potentials in future. First we focus on the nature that marine industry is characterized with experience and know-how like fields as ship's operating, navigation system and port logistics. We show that where some marine systems are running with knowledge and know-how, fuzzy theory will be useful to represent in a systematic way their ambiguity and vagueness. Second we introduce two kinds of applications in marine industry : applications using orthodox methodology and language-based intelligent systems. As a new direction of research, the intelligent system and brain-style computing system based on natural language using fuzzy theory have been suggested, where the ambiguity and vagueness can be handled that is implicated in a language. Recently language-based system using fuzzy theory has been applied to various fields. In this talk, we introduce two language-based intelligent systems : Linguistic instruction-based steering control system of a ship and linguistic route guiding system in ship's navigation. Finally, we introduce some traditional applications in marine industry and discuss a direction for applying fuzzy theory to marine industry.

View full abstract

We have developed a new analysis method of time sequential data based on moving averages. Firstly, we apply the optimal moving average that makes the residual term being independent. Then, we introduce so-called the super moving averages of this optimal moving average, and we estimate underlying potential forces by comparing the differences of these quantities for various time scales [1]. This method has been successfully applied for analysis of market prices. For the check of validity of this method we have applied this method to pure random walks. It is confirmed that the estimated potential forces are very week in such cases. For real market data the distribution is quite different from this theoretical case, sometimes the observed potential forces are clearly much stronger than random cases. In most cases the observed potential functions are nearly symmetric and in such cases we can only predict the speed of diffusion. However, there are cases that the potential function becomes asymmetric when we can predict the averaged direction of price changes.

View full abstract

One of the most fundamental questions about a living system would be ""How can a living creature with finite complexity can survive in an environment with infinite complexity?"" This question has very profound meaning. We'll explain the concept of co-creation system from this viewpoint in this symposium.

View full abstract

Computation by Self-assembling DNA molecule is one of the most important methods of molecular computing. In this method, a DNA super-molecule consists of several short DNA strands (called a DNA tile) self-assembles a large two-dimensional lattice. This process is equivalent to a kind of computation process of one-dimensional cellular automata. Accuracy of the computation is based on the specific hybridization between complementary strands of DNA. Hence hybridization is not deterministic but stochastic process, it is difficult to obtain an aggregate with no error. In this presentation, several methods to reduce the assembly error of DNA tile system, especially methods called Protected Tile Model (PTM) and Layered Tile Model (LTM) which utilize additional DNA substructures to enhance the accuracy of self-assembly will be explained.

View full abstract

The notion of biomimetic machines formulates one paradigm of the latest biologically inspired research, which designates some examples such as biomimetic robots mimicking the configuration and locomotion of biology, intelligent interface mimicking a communication system through human natural voices and gestures, etc. Moreover, the notion is not necessarily equivalent to the bionics as a fusion of biology and electronics, or the biomechanics that fuses biology and mechanics; rather it includes behavior-based robotics in the sense of focusing on the mimesis of intelligent behaviors for biology In this paper, biomimetic perception, biomimetic control, and biomimetic robot, which are subthemes in the biomimetic machines, are explained, mainly referring to the latest results in our laboratory.

View full abstract

Object identification, classification and modelling in unknown or known environments have become an interesting research topic. In this paper, a fusion algorithm combining fuzzy reasoning and neural network is presented for object classification in mobile robot navigation. The imprecise information generated by sonar range finders are filtered and normalized by using Tsukamoto type fuzzy inference system. Then the filtered range information and the pre-processed image data are fused through a back propagation type feedforward neural network to classify the objects in the world environment. Some experimental results show the effectiveness of the proposed methodology.

View full abstract

As a robot control system by voice commands, a fuzzy-coach player system is considered to deal with the fuzzy implication of human natural languages. In any voice commands, it can be easily imagined that the quantity (e.g., the movement distance) desired by a user differs depending on the relative relationship (or situation) between the robot and the environment, even if the user reuses the same command as used before. Therefore, in order to reflect the human intention to determining the operational quantity of the robot, a method for determining an action of the robot, which can learn the command features of the human, is here proposed by introducing a reasoning mechanism composed of a fuzzy-neural network, whenever the movement distance of the robot is determined. The effectiveness of the present approach is illustrated through some experiments for a block piling task.

View full abstract

For constructing a central pattern generator (CPG) for articulated robots, the Matsuoka oscillator is assumed here to be used as a fundamental oscillator. In particular, we discuss how to connect each oscillator to construct a stable CPG network, from the view point of the symmetry of the network, the inhibitory and/or the excitory connections, etc. Some CPG networks using such oscillators are designed for one leg model in a human-like robot.

View full abstract

This paper demonstrates that Neural Networks can be used effectively to compensate for incorrectly selecting gains of the PID controller. Selecting appropriate the gains of the PID controller is difficult and is time consuming. If the gains of the PID controller are not suitable, then the performance of the PID becomes poor. Therefore, selecting the PID gains is very important. In this paper, the Neural Network (NN) is used as a ""plug and play"" module for the PID controller. When the PID gains are incorrect, the NN take over the controller, otherwise the NN doesn't operate. A simple structure, fast computing, and on-line training NN has been designed for that purpose. We applied the proposed controller to the magnetic levitation test bed and the experimental results showed that using the proposed NN in the PID controller can improve the performance of the PID controller.

View full abstract

In this paper it is presented a mixed control approach for stabilising a Furuta pendulum in upright position. The full Monty controller consists partly of a state feedback controller and partly of a Lyapunov controller. A fuzzy controller takes the decision to vary between of the structures.

View full abstract

This paper proposes an aeration control method on advanced wastewater treatment processes for stability of a control system. A dissolved oxygen meter and ASM1 (Activated Sludge Model No.1) suggested by IWA (International Water Association) are utilized in the method. The method is compare with an aeration control method with an ammonium nitrogen meter. The simulation results show the effectiveness stability of the proposed method.

View full abstract

This paper is concerned with output feedback stabilization with guaranteed cost of Takagi-Sugeno fuzzy systems with immeasurable premise variables. When we consider Takagi-Sugeno fuzzy systems, the selection of premise variables plays an important role. If the premise variable is the state of the system, then a fuzzy system describes a wide class of nonlinear systems. However, the state is not measurable in the output feedback control problem. In this case, a control design of the underlying nonlinear system based on its fuzzy system description is difficult. In this paper, we consider a problem of the output feedback stabilization with guaranteed cost. In this problem, the premise variable may not be measurable. Thus we formulate the output feedback stabilization with guaranteed cost problem of fuzzy systems as a robust stabilization of uncertain systems. Numerical examples illustrate our theory.

View full abstract

In this paper, we propose an action rule extraction technique from RoboCup soccer log files. Our technique extracts action chains from log files to determine the consequent part of each action rule. From results of computational experiments, we show that the effectiveness of our technique for generating action rules automatically from log files.

View full abstract

OZED is an educational virtual soccer robot construction kit that enables users creating their own soccer agent which is able to connect official RoboCup soccer simulation servers. A new kind of user interface system KONOHEN fuzzy was introduced and developped. That is a resoning with fuzzy set on multidimensional arbitrary shape.

View full abstract

This paper proposes a framework for acquiring a low-level behavior of a soccer agent. The task of a learning agent is to mimic the behavior of a target agent with a well-trained behavior. Neural networks are used to represent the behavior of the target agent. In order to obtain a set of training data, we convert game logs of the target agent into a set of input-output pairs for the neural networks. We show the effectiveness of the proposed framework through the computational experiments.

View full abstract

What is {\em teamwork} in multi-agent systems? It is not trivial to give a suitable definition of {\em teamwork} for teams of agents working together, because agents do not have innate goals, purposes, or instincts. This difficulty becomes clear when we consider {\em teamwork} in soccer-playing teams of agents in which all agents share the same goal and purpose. In order to approach to suitable definitions, I investigate {\em teamwork} from several point of view and try to figure out the possible definitions.

View full abstract

A team formation is a basic and important factor that determines a characteristic of a team strategy in the RoboCup Soccer Simulation. In previous researches, Situation Based Strategic Position(SBSP) has been used as a standard formation mechanism, because this technique is very simple and has high performance. However, we must consume huge manpower to optimize the formation using SBSP. To solve this problem, we propose a GUI tool to construct a team formation by human's intuitive operations. Our proposal method enables us to make a flexible team formation easily and quickly. To compare our method with SBSP, we had testees construct a team formation by these two methods. As a result, FormationEditor showed a higher performance than SBSP.

View full abstract

Interactive Evolutionary Computation (IEC) is one of the effective methods for optimization problems which is difficult to formulate the evaluation function such as human sensitivity. This method, however, gives a big burden to the user because he/she himself/herself has to evaluate the candidates of solutions a lot of times. This paper employs fitness inference method to reduce the burden for evaluation. When fitness inference method is applied to IEC, it can be a problem that the evaluation criterion for user has changed with the passage of time and/or impression given by the effects of other candidates. Because the fitness inference method infers the fitness values of candidates using the information of the actually evaluated solutions and their fitness values in previous generations. This paper proposes the fitness inference method with varying the number of actual evaluation for candidates based on the accuracy of inference for fitting the change of evaluation criterion of user. The proposed method can reduce the burden of evaluation following the user's evaluation criterion. This paper applies the proposed method to hearing aid adjustment support system by IEC and investigates the effectiveness of this method.

View full abstract

Reinforcement learning is a promising approach to realize intelligent agent such as autonomous mobile robots. In order to apply the reinforcement learning to actual sized problem, the ""curse of dimensionality"" problem in partition of sensory states should be avoided maintaining computational efficiency. The paper describes a hierarchical modular reinforcement learning that memory based modular Profit Sharing learning algorithm is combined with Q Learning reinforcement learning algorithm hierarchically in multi-agent environment. By using the hierarchical modular reinforcement learning, this research aims at learning the behavior of the agent in the pursuit problem, i.e. reinforcement learning problem in the multi-agent environment. Through numerical experiments, we found that the proposed method has good convergence property of learning compared with the conventional algorithms.

View full abstract

Optimization requirements often include a finding various solutions and a searching under muti-objective situations. A maintaining diversity of individuals is one of the effective approaches to meet the requirements, when we use Genetic Algorithms (GAs) for these optimization. Our research aims to maintain the diversity. We propose a new selection scheme for maintaining the diversity and apply the selection to simple GA (sGA). In our selection, the individuals are classified by Fuzzy c-Means. Accordingly, several clusters come up and each of the individuals gets a membership value for each of the clusters. The proposed method selects individuals based on both the fitness values and the membership values. We also discuss about behavior and search capability of the GA with the proposed selection method via some simulations. Based on results of the simulations, we were able to find out that the GA makes the individuals widely distributed in a solution space.

View full abstract

Generally, as for Genetic Algorithms (GAs), it is not always optimal search efficiency, because genetic parameters (crossover rate, mutation rate and so on) are fixed. For this problem, we have already proposed Fuzzy Adaptive Search Method for GA (FASGA) that is able to tune the genetic parameters according to the search stage by the fuzzy reasoning. On the other hand, in order to improve the solution quality of GA, Parallel Genetic Algorithm (PGA) based on the local evolution in plural sub-populations (islands) and the migration of individuals between islands has been researched. In this research, Fuzzy Adaptive Search method for Parallel GA (FASPGA) combined FASGA with PGA is proposed. Moreover as the improvement method for FASPGA, Diversity Measure based Fuzzy Adaptive Search method for Parallel GA (DM-FASPGA) is also proposed. Computer simulation was carried out to confirm the efficiency of the proposed method and the simulation results are also reported in this paper.

View full abstract

This paper proposes a gene linkage identification method during local search for permutation problems. Linkage is defined as the tightness of loci in a chromosome. In the literature, many studies on linkage are devoted to binary-coded chromosomes. We have proposed an idea to identify the gene linkage in permutation problems where a chromosome is represented by a permutation of different alleles. We extend our proposed method to identify gene linkage during local search. Using our extension, we improve the performance of genetic local search to find near-optimal solution in flowshop scheduling problems.

View full abstract

We studied ways of extracting important information from natural language processing article abstracts. While many kinds of information extraction, such as from newspapers, web materials, or biology article abstracts, already exist, no papers have been written about extracting natural language processing information. We hope that this study will be useful for natural language processing researchers. We defined nine categories that contain important expressions for natural language processing article abstracts. We constructed a method of extracting these expressions by using machine learning methods. Our method extracted these expressions with an F-measure of 0.67. When we considered partially correct expressions to be correct, the F-measure increased to 0.73. In particular, important expressions belonging to four categories (Accuracy, Field, Language, and Name) were automatically extracted at a high F-measure (about 0.8 to 0.9) using our method. We next constructed various kinds of visualization tools using important extracted expression. They can display the abstract of a paper highlighting extracted important expressions with color, display abstracts in a table including extracted inpotant expressions in each row, and make a graph indicating the distribution and trend of papers including important expressions for each category.

View full abstract

We constructed a system for this study to automatically extract numerical pairs from documents related to a certain topic and display them in graphs. It first extracts two units and one item expression from documents, it then extracts numerical pairs from sentences including the two units and the item expression, and it finally arranges the pairs and displays them in graphs. When we judged the extraction of a correct graph as the top output in the experiments to be correct, our best system accuracy was 0.2222 in Evaluation A and 0.4444 in Evaluation B. When we judged the extraction of a correct graph in the top five outputs to be correct, the best accuracy rose to 0.3889 in Evaluation A and 0.5000 in Evaluation B. (In Evaluation A, a graph where 75\% or more of the points was correct was judged to be correct and in Evaluation B, a graph where 50\% or more of the points was correct was judged to be correct.) Our system is convenient and effective because it can output a graph that includes numerical pairs at these levels of accuracy when only given a set of documents as input. We also showed some graphs that were automatically constructed by our methods and taught us many interesting things in this paper.

View full abstract

This paper proposes a new interface, equiped with knowledge that supports to achieve user goal, that can operate plural application software simultaneously with natural language. In order to decide which application software should be operated by given natural language instruction, we have poroposed dictionaries to identify the concrete value of how much a certain vocabulary is related to each application software. By this, it has become able to operate plural application software with one interface by referring to this value. We show an example of operating plural application software. In this stuy, we used MS Word, MS Excel, and MS Outlook as the target application software.

View full abstract

In this paper, we have discussed intelligent information providing with an example that news articles about the stock prices are summarized based on the representation of a 2D chart on the stock prices. We used MuST corpus as news articles to be summarized, which is an annotated corpus for extracting easily trends information in the corpus. We have associated the MuST corpus with numerical data of the stock prices, and proposed a method to provide people with a summarized text about the news articles based on the representation of the 2D chart of the prices.

View full abstract

A large amount of molecular level information on biological phenomena, such as embryogenesis, cell proliferation, cell differentiation, signal transduction, transcriptional regulation, cell response, etc. is rapidly increasing because of the progress of genome science in recent years. In order to understand high-order biological phenomena, it is necessary to bridge the gap between the amount of information in the molecular level and the amount of information in the cell functional level. However, the biological knowledge in the cell functional level has not been systematized well. We have been seeking a method to systematically describe and understand cellular behaviors in morphogenesis. Various morphogenetic processes were manually selected and described according to the following format: ""cell(s)=(subject)"" ""does something=(output matter)"" ""under a certain condition=(input matter)"". CBDB is a biological database on cell behaviors in the format with images and movies which indicate To sophisticate and make use of the CBDB, we propose an AI-based approach for more multiple searching of instances and knowledge discovery from the CBDB. As the first step, we attempted to construct the ontology for understanding morphogenetic phenomena. The ontology will also facilitate clearer definition of terms and concepts, and knowledge sharing among communities of various fields. In this study, we introduce the CBDB and the prototype of ontology-based information retrieval system.

View full abstract

In this paper, parked vehicles license plate recognition using image frame-difference and template matching is proposed. Image frame-difference is used to extract the vehicle area, reducing the license plate search domain, thereby reducing the total search time. The processes of expansion and reduction are then applied to reduce the noise in the extracted plate domain. Based on the data from 83 vehicles, the most likely plate area is then decided. License plate recognition is carried out using template matching. For each possible character, three templates are prepared. We then performed computer simulations to show the effectiveness of our method. From the result, it was confirmed that the preprocessing method was effective.

View full abstract

Morphing is a process of smoothly transforming from one shape into another. In this paper, we present a feature-based method of employing mesh morphing to compute two homeomorphic facial meshes for animating 3D facial expressions in virtual worlds. We divide the method into five steps, including head decomposition, mesh parameterization, feature specification, vertex correspondence, and point interpolation. Based on sparse sets of user specified feature points, a semi-automatic ripple searching algorithm is proposed to cope with the problem of establishing a one-to-one correspondence between source and target meshes. After the operation of correspondence establishment, we employ Bezier interpolation with four data control points to transform from the source mesh to the target mesh. Finally, three pairs of facial meshes with surface textures are demonstrated to produce a series of facial expressions by using the proposed method.

View full abstract

An automatic 3D pair-wise registration method based on an optimal segmentation and a local character tensor is proposed. It consists of two characters, i.e., the optimal segmentation to realize the automatic processing and the local character tensor to improve the matching probability. Non-structured 3D datasets of a rabbit doll are tested on a PC with Intel Pentium M 1.60GHz and 512MB memory. Experiment results show the proposed method increases the matching probability average for 18.5% and decreases the computational time average for 21.6 second compared to those of the conventional spin image method.

View full abstract

This paper discusses a vision-based teleoperation system for human-friendly interface of a mobile robot used under office automation floors. We proposed a navigation method based on human simple commands. Furthermore, we proposed a perceptual system for estimating the self-location and action system for controlling the posture of the robot. This paper proposes a human-friendly information display system to improve the operability of the robot navigation. Finally, we show several experimental results of the proposed method.

View full abstract

This paper proposes the use of gradient orientation information, instead of traditional image information such as intensities and gradient vectors, for estimating motion in image sequences. Gradient orientation information may be exploited with unit gradient vectors that are invariant to image intensities to a great extent. By replacing image intensities or gradient vectors with the unit gradient vector, the robustness of traditional motion estimation techniques to changes in lighting conditions can be significantly improved. We describe two motion estimation techniques based on gradient orientation information. The proposed techniques work robustly regardless of varying image intensities. In addition, they show an improved tolerance to low image contrast, and the so-called aperture problem that makes existing motion estimation techniques ineffective.

View full abstract

An algorithm for the elicitation of single knots from the utility curve of real decision makers is proposed. The preferential equations of four utility elicitation techniques, namely probability equivalence, certainty equivalence, lottery equivalence and uncertain equivalence are derived from a generic theoretical setup of elicitation methods and a practical algorithm is elaborated for the application of the uncertain equivalence method to both monotonically increasing and decreasing preferences. The algorithm determines uncertainty intervals for the roots using triple bisection. Two example cases are discussed, and the preferences of the decision maker are elicited in one of them through a combination of uncertain equivalence - lottery equivalence algorithm.

View full abstract

In this paper, we examine how to reach Pareto optimal equilibrium in Double-bind Prisoner's Dilemma game (hereinafter referred to as DbPD game). DbPD game are made by adding dominated strategy to 2 by 2 prisoner's dilemma game (Wada and Suzuki (2006)), and the problem of playing DbPD game or not also becomes a dilemma. Though our agents are all set as rational, selfish, reinforcement-learning based model agents, we discover that they learn to avoid using Nash equilibrium in early round. After finishing to learn that, then they start to learn cooparating. Through these process, our agents are possible to contribute each other in DbPD game.

View full abstract

In pursuit of business scope expansion or global implementation of business strategies, corporates always look for the opportunities of overseas project investment. However, during the process of overseas project investment, they have to face more uncertainties comparing with investing domestically, for example, rules and regulations, cultural background, and currency fluctuation of the target country etc. Consequently, it is essential for corporates to undergo a full-scaled risk evaluation before the practices of overseas project investment, so as to properly manage all the risk factors, set control strategies or enhance the success probability of project investment. Therefore, the means to build a risk evaluation model for overseas project investment is an extremely important issue for the corporates before practically undergo an overseas project investment, so as to consider all kinds of risk factors accurately and effectively. This study adopts related literatures published locally and internationally, and analyses into the risk factors provided by corporates having related experiences. With the intention of providing benefits to corporates in times of overseas investment, fuzzy analytic hierarchy process (FAHP) and fuzzy multi-criteria decision making (FMCDM) are then employed to build a scientific and systematic risk evaluation model.

View full abstract

The paper proposes three types of neural models, namely the fixed neural model (FNM), the growing neural model (GNM) and the evolving neural model (ENM) and their respective learning algorithms, based on different types of available data. The off-line learning algorithms of the FNM and GNM use batch-type (off-line) data sets, while the ENM learning can be performed in real time, by using ""endless"" data stream. It is shown in the paper, that the GNM is more flexible than the FNM and faster in off-line learning, while the ENM is a very convenient tool for analysis of real-time data, with its ability to analyse time-varying machine operations. Test examples as well as real data from a diesel engine of a hydraulic excavator are used in the paper to demonstrate te applicability and the features of the proposed neural models.

View full abstract

Kohonen's Self-Organizing feature Map (SOM) is a method to obtain a topology-preserving mapping from high-dimensional invisible feature space to two or less dimensional visible space. For obtaining a good mapping, the user has to set up a good combination of parameters, type of lattice, number of neurons and neighboring radii varied with iteration etc.. This paper proposes a new method to obtain a mapping without any setting of parameters except for the termination condition. The method starts with six neurons. And the method moves positions of neurons in the visible space and generates new neurons in accordance with the data distribution in the feature space.

View full abstract

One way to understand a protein's function is to infer from its structure. However, it is not easy to obtain the three-dimensional structure directly. So to predict substructures like alpha helices, beta sheets, and coils from the protein sequence is the first step to decode the mystery of a protein. It is straightforward and was investigated, but improvement is still needed in this problem. In this paper, the gamma neural model for context analysis is used. Also, a new encoding method for the protein sequence is proposed and discussed. The results are compared with the traditionally-used time-delay neural networks and another encoding method in the field. The gamma neural model not only can reduce both space complexity and time complexity but also improves the overall prediction accuracy. The proposed encoding method, on the other hand, shows its usefulness in improving the prediction rate of beta sheets.

View full abstract

In this paper we describe an autonomous mobile robot whose sensory-motor connection were made by a three-layered spiking neural network (SNN) with only one hidden-layer neuron that makes synaptic contacts on motor neurons with synapses having spike timing-dependent plasticity (STDP) and presynaptic modulation, and we analyzed the roles of STDP for the autonomous behavior in environment. STDP is known as a kind of Hebbian rules with which the synaptic weights are increased or decreased in accordance with the relative timing of pre- and postsynaptic action potentials. An excitatory synapse may become an inhibitory one, or vise versa, in the STDP. We used one of the SNN models, called the spike response model (SRM), in which the neurons generate spikes, and a spike at presynaptic site generates a delayed, prolonged post synaptic potential (PSP). The postsynaptic neuron fires when the sum of PSPs becomes over a threshold. Once a neuron fires, it goes into a refractory period during which a larger input is necessary to generate the following spike. We considered a mobile robot with left and right front proximity sensors, and left and right wheels driven by independent motors. Each sensor value was converted to the probability of spike generation by the sensory neuron. The outputs from left and right sensory neurons were connected to one hidden-layer neuron with fixed synapses. The output of the hidden-layer neuron makes synaptic contacts on left and right motor neurons, and these synapses have the properties of STDP in regard to the presynaptic modulation signals from sensory neurons. The spiking rates were converted to analog values to drive the corresponding motors. We implemented this SNN in a miniature mobile robot Khepera. In a given environment, it gradually organized the weights and acquired the navigation and obstacle-avoidance behavior. Modeling and experimental data analyses showed that one hidden-layer neuron is sufficient for the task, and that such SNN is suitable for the navigation of mobile robots.

View full abstract

In this paper, we try to estimate Japan's cabinet approval ratings by using neural networks. Though there are a number of studies on approval rating estimation, little attempts have been made to apply neural networks to estimate approval ratings. Especially, in Japan, no attempts have been made to infer approval ratings by neural networks. Thus, this is the first attempt to use neural networks for approval estimation in Japan. Neural networks may show better performance for this complex problem, because approval ratings are affected by many complex factors that cannot be dealt with by conventional statistical methods. Experimental results show that neural networks have much better performance than that obtained by the standard regression analysis in terms of training and testing errors.

View full abstract

This paper describes a human-friendly interface based on hand gesture. Using two USB pan-tilt cameras, this system enables the user to operate home appliances with hand pointing and ten hand motions free from the user's position. The main contribution of the system lies in its user-friendly initialization methods including interaction between the user and the system. The user can initialize or modify the map describing the location of home appliances to control simply by hand gesture. Embedded feedback mechanism enhances the overall performance by adjusting user's miss pointing.

View full abstract

The studies on automatic ship collision avoidance system, which have been carried out in the last 10 years, are facing on new situation due to newly developed high technology such as computer and other information system. It was almost impossible to make it used in real navigation field 3-4 years ago because of the absence of any tool to give other ship's information, however recently developed technology suggests new possibility. This study is carried out to develop the automatic ship collision avoidance support system which considers ship's manoeuvrability into it's collision avoidance algorithm. One of the important part in ship collision avoidance system is collision decision module which can calculate collision risk with other ship's and act properly to avoid the situation. Many of previous researches are using present ship's dynamic data such as present speed, position and course to calculate collision risk. However when a ship commences avoidance action, the real situation is quite different with one that has been estimated by the ship's initial data due to the ship's manoeuvring characteristic. Therefore it is better to take into account ship's manoeuvring charateristic from the stage of collision decision in ship collision avoidance system. In this study, these effects are included in the developed system. The proposed system are vilified its usefulness in numerical simulation environments.

View full abstract

Decision Tree Algorithm Improved with a Time-weighted Entropy

View full abstract

Ship's collision avoidance is a skill that Masters of merchant marine vessels have acquired through years of experience and that makes them feel at ease to guide their ship out from danger quickly compared to inexperienced officers. Case based reasoning (CBR) uses the same technique in solving tasks that needs reference from variety of situations. CBR can render decision-making easier by retrieving past solutions from situations that are similar to the one at hand and make necessary adjustments in order to adapt them. In this paper, we propose to utilize the advantages of CBR in a support system for ship's collision avoidance while using fuzzy algorithm for its retrieval of similar navigational situations, stored in the casebase, thus avoiding the cumbersome tasks of creating a new solution each time a new situation is encountered. There will be two levels within the Fuzzy-CBR. The first level will identify the dangerous ships and index the new case. The second level will retrieve cases from casebase and adapt the solution to solve for the output. While CBR's accuracy depends on the efficient retrieval of possible solutions to be adopted from stored cases, fuzzy algorithm will improve the effectiveness of solving the similarity to a new case at hand.

View full abstract

Our starting point is the multiplicative utility function which is extensively used in the theory of multicriteria decision making. Its associativity is shown and as its generalization a fuzzy operator class is introduced with fine and useful properties. As special cases it reduces to well-known operators of fuzzy theory: min/max, product, Einstein, Hamacher, Dombi, drastic. As a consequence, we generalize the addition of velocities in Einstein's special relativity theory to multiple moving objects. Also, a new form of the Hamacher operator is given, which makes multi-argument calculations easier. We examined the De Morgan identity which connects the conjunctive and disjunctive operators by a negation. It is shown that in some special cases (min/max, drastic,Dombi) the operator class forms a De Morgan triple with any involutive negation.

View full abstract

There are relatively few Fuzzy Rule Interpolation (FRI) techniques can be found among the practical fuzzy rule based applications. On one hand the FRI methods are not widely known, and many of them have limitations from practical application point of view, e.g. can be applied only in one dimensional case, or defined based on the two closest surrounding rules of the actual observation (which requires a time consuming search in case of multidimensional rule base spaces). On the other hand enabling the application of sparse rule bases the FRI methods can dramatically simplify the way of fuzzy rule base creation, since FRI methods can provide reasonable (interpolated) conclusions even if none of the existing rules fires under the current observation. These methods can save the expert from dealing with derivable rules and help to concentrate on cardinal actions only, and hence simplify the rule base creation itself. Thus, compared to the classical fuzzy compositional rule of inference (CRI), the number of the fuzzy rules needed to be handled during the design process, could be dramatically reduced. In this paper, among the brief structure of several FRI methods, a simple and quick FRI method ""FIVE"" will be introduced in more details, and for demonstrating the benefits of the interpolation-based fuzzy reasoning as systematic approach, the construction of a fuzzy rule base through a simple example will be also discussed.

View full abstract

A formation of the shortest path from a nest to a food area was investigated by using small ant robots. Ants communicate by a pheromone in nature. When an ant discovers food, it goes back marking a pheromone on its way to the nest. Its fellows follow the trace of the pheromone and then they also mark the pheromone on their way to the nest after arriving at the food area. At the early stage, the route is a complicated line. A lot of ants' coming and going optimize the route from the nest to the food area. We propose the new pheromone field that has an important feature of real pheromone:i.e., evaporation. Under the condition of the existence of the obstruction on the way to the food area, the experiments of path optimization were carried out by using both the ant robot and the pheromone field. The experiments demonstrated that our ant system was applicable to the optimization problem.

View full abstract

In our research group, an assist method for human's operation of electric-powered wheelchairs has been investigated. The purpose of researches is to make powered wheelchair intelligent and to develop a mobility aid for people, who find it difficult or impossi-ble to drive a conventional electric powered wheelchair. On some prototype of our group, a neural network produces an obstacle avoidance function. In previous researches, we found that by varying connection weights of the neural network according to the condition of ob-stacles in the neighborhood of the wheelchair, the obstacle avoidance function is improved. In this research, we discuss the adjustment ability of a neural network with variable type connection weights based on numerical studies.

View full abstract

The paper proposes a patient robot for training and evaluation of nursing skill of student nurses. The patient robot will be used for not only training but also objective evaluation of nursing action for improvement of nursing skill. For example, looking at the patient face in injection is important so that the patient feels safe. As a basic experiment, small CCD cameras are embedded into the eyes of a head of a simulation model in order to evaluate whether the nurse is looking at the patient face. Our method employs Genetic Algorithm (GA) and unprocessed input image to recognize frontal face of nurse in real time by model-based matching method. Moreover, HSV color system is adopted to improve the tracking performance of the recognition against noise while keeping the real-time nature. The experimental results show the method can evaluate injection skill in nursing by detection of frontal face of nurse.

View full abstract

Log-Polar transform is an approximation model of human vision. If using it to recognize human faces, a reference point is necessary, and it influences the recognition accuracy directly. However, it is difficult to extract an appropriate reference point from an image automatically. This paper proposed a face recognition method that finds out the center of a face as a reference point of Log-Polar transform automatically. With this method, it is easier to recognize the faces by using Log-Polar transform.

View full abstract