Search

Many researchers are concerned with the analysis of human intention, which is based on decision-making. However, understanding someone's intentions can sometimes be difficult because it is related to many patterns that are dependent on individual situations. In order to analyze intention, we used a brain-mapping system to study the human brain. We call this the brain-computer interface (BCI) system. Earlier studies of BCI were based on visual stimuli. Using auditory stimuli, however, is still in its infancy in the field of research. In this paper, we proposed an auditory paradigm using the auditory steady-state response (ASSR) and canonical correlation analysis (CCA) to improve the accuracy of selective attention. It was made by our own experimental environment using the event-related potentials as P300. We describe a paradigm that allows a user to make a decision based on intention. We studied an auditory paradigm-based brain computer interfaces for intention recognition and how we can use it in BCI systems.

In previous paper, we proposed the novel method of nonlinear unsupervised feature classification for EEG (Electroencephalography) signal based on HS (Harmony Search) algorithm. Using this method, we could convert classification problem into finding the smallest sum of Euclidean distances between vectors belonging to each class. Therefore the performance of proposed method was influenced by the performance of optimization algorithm. In this paper, to compare efficiency and performance of various heuristic algorithm for this method, we applied three different heuristic optimization algorithm, HS, PSO (Particle Swarm Optimization), and DS (Differential Search). For the simulation, we used EEG signal data from BCI Competition IV Dataset I. Two class data from two subject with 100 Hz sampling rate were used. For feature extraction, common spatial pattern algorithm was used. In conclusion, the fastest algorithm was HS algorithm with about 4.4 seconds of an average computational time, the algorithm with best classification rate was also HS algorithm and the average classification rates of subject `f' and `g' were 84.08 % and 81.95 %. The slowest heuristic algorithm was PSO algorithm with about 7.5 second in an average computational time, and the worst average classification rate was 57.27 % from subject `g' with PSO algorithm. We could draw a conclusion that the best algorithm for proposed classification method was HS algorithm.

Classification of limbs motion intention based on electroencephalogram (EEG) is one of the major subjects of brain-computer interface (BCI). EEG signals contained not only noise artefacts, but also variations within and across sessions. To reduce this unnecessary information in EEG signal processing, we applied an extension version of common spatial pattern (CSP). The original CSP has been widely used to extract relevant spatial features from EEG signals. However, one of the CSP limitations is occurred when it concentrated on artifacts and variances sources, which exceeded the variance of endogenous component of the brain, instead of extracting the sources that provide subject's intention. Therefore, we proposed to add more parameter for separating invariance and variance in EEGs. To reduce the variant parts of signal in EEGs, we applied invariant common spatial pattern (iCSP) to maximize one class in the same time minimize other class and invariant sources. Based on neurophysiological knowledge, we made some strong weights to special electrodes related to our experimental paradigm. The raw EEG signals were recorded by auditory stimuli with 3 different frequencies cued for real hand movement and ignored noise. The right or left actual hand movements-based EEG signal states were classified by using linear discriminant analysis (LDA).

Brain computer interface (BCI) system allows users to direct interact with the surrounded environment just a blink of their thought. In doing this, the most relevant informative from the electroencelography (EEG) signals need to be extracted from the electrodes of the scalp. Neurophysiology studies have proved that the power density from corresponding electrodes and frequency could identify imagination of the left or right hand movement. They also proved that these feature vary strongly from one subject to another. These spatial-time-frequency components are the keys to unlock the optimal features from the large space of these power density features. In this paper, we proposed the optimal feature extracting method from the basic power density of EEG signal. At first, the all EEG signal from the electrodes were filter using both spatial and temporal filters to enhance the signal to noise ratio of EEG. Then, the time-frequency features were extracted using short-time Fourier transform (STFT) and average power in sub-window band. Genetic algorithm was applied to search for the optimal features. In our simulation, we used the dataset from BCI competition III, IV and the data experimented in our laboratory. To ensure the improvement of our proposed feature extraction method, we applied the extracted feature into the support vector machine.

Support Vector Machine maximizes a margin between two groups. Variance-considered machine improves SVM to align hyper plane according to two classes' variance and prior probability to reduce the error rate. There is probabilistically imprecise things those data classified by VCM. In this paper, we introduce the VCM and try to propose a concept that is to confer reliability estimated by Mahalanobis distance upon data separated by VCM.

A motor imagery related electroencephalogram (EEG) feature classification technique through the time-series prediction based on the adaptive neuro-fuzzy inference system (ANFIS) is presented for neural computation applications. We descries a method for classification of EEG using optimized ANFIS and the proposed method was focus on the validation of the Harmony Search algorithm based optimization procedure for ANFIS. The motor imagery related EEG signals recorded from subjects performing left, right hand and foots motor imagery. From this signal, features obtained from the difference of multiresolution fractal feature vectors between the predicted and actual signals by using time-series prediction technique. In order to optimize the ANFIS, Harmony Search algorithm is sufficiently adaptable to allow incorporation of other training techniques like feed-forward and gradient descents. In this paper, the proposed technique is employed to simulate the three types of motor imagery (left, right hand, right foots) EEG signals evaluation data which were used as input patterns of the optimized ANFIS classifier.

Music is the pattern of sounds produced by people singing of playing instruments. Thus, the music is made by people. Music composer, will be planted in the human emotions, put the listener is able to feel a similar sentiment. Especially, interesting from our point of view is that music analysis, the emotions of the people can be analyzed. This paper focuses on the analysis of auditory cortex active status by music genre. In this paper, the musical stimuli in EEG signals by amplifying the corresponding reaction to the averaging method, ERP(Event-Related Potentials) experiments based on the process of extracting sound methods for removing noise from the ICA algorithm to extract the tone and noise removal according to the results are applied to analyze the characteristics of EEG. In addition, drawing on LORETA(Low Resolution Brain Electromagnetic Tomography)program have attempt to design location of electrode in brain. And finally, using EEGLAB, the relationship of music and the auditory cortex active status could be concluded.

For our present life anytime, anywhere access to the network can communicate with the ubiquitous computing. it is essential to human life. We should be able to agree that communication will be enabled. For our present life, anytime, anywhere access to the network can communicate with the ubiquitous computing. Such as the ubiquitous era approached, interaction between the user and the computer has become an important issue. In this paper we use EEG signals to extract the user's intention recognition data, which the Mirror Neuron System Based on HMMs and GMMs to model the convergence of the hybrid model is proposed. This is based on a kind of biological signals using EEG signals to the user's intention recognition techniques have been studied. In addition, EEG signals is generated based on the model, using the user intention recognition method have been studied. The proposed model will be applied in the field of neuro robotics.

Usually, the video based object tracking deal with non-stationary image stream that changes over time. Robust and Real time moving object tracking is a problematic issue in computer vision research area. Multiple object tracking has many practical applications in scene analysis for automated surveillance. If we can track a particularly selected object in an environment of multiple moving objects, then there will be a variety of applications. In this paper, we introduce a particular object tracking in an environment of multiple moving objects. When tracking, we need to analyze video sequences to track object in each frame. In this paper, we use a differential image of region-based tracking method for the detection of multiple moving objects. In other to ensure accurate object detection in unconstrained environment, we also use a method of background image update. There are problems in tracking a particular object through a sequence of video. It can't rely only on image processing techniques. Thus we solved these problems using a probabilistic framework. Particle filter has been proven to be a robust algorithm to deal with the nonlinear, non-Gaussian problems. In this paper, the particle filter provides a robust object tracking framework under ambiguity conditions and greatly improved estimation accuracy for complicated tracking problems.

The mobile robot which accomplishes a work in explored region does not know location information of surroundings. Traditionally, simultaneous localization and mapping(SLAM) algorithms solve the localization and mapping problem in explored regions. Among the several SLAM algorithms, the EKF (Extended Kalman Filter) based SLAM is the scheme most widely used. The EKF is the optimal sensor fusion method which has been used for a long time. The odometric error caused by an encoder can be compensated by an EKF, which fuses different types of sensor data with weights proportional to the uncertainty of each sensor. In many cases the EKF based SLAM requires artificially installed features, which causes difficulty in actual implementation. Moreover, the computational complexity involved in an EKF increases as the number of features increases. And SLAM is a weak point of long operation time. Therefore, this paper presents a symmetrical model based SLAM algorithm (called M-SLAM).

One of the critical components of localization is the robot's ability to locate itself on a partially map. This becomes challenging when the robot experiences positioning error, does not have an external positioning device, nor the luxury of engineered landmarks placed in its free space. This paper presents a new method for localization. When the mobile robot works, The robot from within environment must presume location of oneself accurately. For this, The location recognition algorithm is required tracking location from start point and recovering lost location. In this paper, we propose location recognition algorithm using advanced particle filter for tracking location from start point and recovering lost location. As a result, it can increase robustness and accuracy of localization. The performance of the proposed algorithm was verified by experiments.

In this paper, novel methods for facial emotion recognition in facial image sequences are presented. Our facial emotional feature detection and extracting based on Active Appearance Models (AAM) with Ekman’s Facial Action Coding System (FACS). Our approach to facial emotion recognition lies in the dynamic and probabilistic framework based on Dynamic Bayesian Network (DBN) with Kalman Filter for modeling and understanding the temporal phases of facial expressions in image sequences. By combining AAM and DBN, the proposed method can achieve a higher recognition performance level compare with other facial expression recognition methods. The result on the BioID dataset show a recognition accuracy of more than 90% for facial emotion reasoning using the proposed method.

If we want to recognize the emotion via the face-to face interaction, first of all, we need to extract the emotional features from the facial image by using a feature extraction method. The active appearance model (AAM) is a well-known method that can represent a non-rigid object, such as face, facial expression regions as emotional features. And then we need to classify the emotional status reliably, robustly. Bayesian Network is a probability based classifier that can represent the probabilistic relationships between sets of facial features. So, in this paper, our approach to facial feature extraction lies in the proposed feature extraction method based on combining AAM with Facial Action Coding System (FACS) for automatically modeling and extracting the facial emotional features. To recognize the facial expression, we use the Dynamic Bayesian Networks (DBNs) for modeling and understanding the temporal phases of facial expressions in image sequences.

In the facial image, emotions are most widely represented with eye and mouth expressions. If we want to recognize the human's emotion via the facial image, we need to extract features of the facial image. Active Shape Model (ASM) is one of the most popular methods for facial feature extraction. Regarding the traditional ASM depends on the setting of the initial parameters of the model, in this paper we propose a facial emotion recognizing method based on ASM and Bayesian Network. Firstly, we obtain the reconstructive parameters of the new gray-scale image by sample-based learning and use them to reconstruct the shape of the new image and calculate the initial parameters of the ASM by the reconstructed facial shape. Then reduce the distance error between the model and the target contour by adjusting the parameters of the model. Finally get the model which is matched with the facial feature outline after several iterations and use them to recognize the facial emotion by using Bayesian Network.

In this paper, we used particle filter to motion estimation algorithm on real-time for mobile surveillance robot. Particle filter based on the Monte Carlo's sampling method, be used Bayesian conditional probability model which having prior distribution probability and posterior distribution probability. By using particle filter, it can be possible to tracking and estimating robustly for object's motion and movement. Also most of the initial probability density was set to define or random manually. Proposed method in this paper, however, using the Sum of Absolute Differences (SAD) is to take the initial probability density. Therefore, by using a particle filter to the object tracking system, it can be configured more efficient.

Even though the genetic algorithm is known to be a very effective method to solve the global minimization problem, it needs much time (a large population size and a large number of generations) for a reliable answer and thus it seems to be inadequate for on-line performance. We propose a population feedback GA scheme. we show the effectiveness of our scheme by finding an observer for the discrete-time nonlinear autonomous systems with simulations.

We addresses the issue of expressive face modeling using an advanced active appearance model for facial emotion recognition. We consider the six universal emotional categories that are defined by Ekman. In human face, emotions are most widely represented with eyes and mouth expression. If we want to recognize the human's emotion from this facial image, we need to extract feature points such as Action Unit(AU) of Ekman. Active Appearance Model (AAM) is one of the commonly used methods for facial feature extraction and it can be applied to construct AU. Regarding the traditional AAM depends on the setting of the initial parameters of the model and this paper introduces a facial emotion recognizing method based on which is combined Advanced AAM with Bayesian Network. Firstly, we obtain the reconstructive parameters of the new gray-scale image by sample-based learning and use them to reconstruct the shape and texture of the new image and calculate the initial parameters of the AAM by the reconstructed facial model. Then reduce the distance error between the model and the target contour by adjusting the parameters of the model. Finally get the model which is matched with the facial feature outline after several iterations and use them to recognize the facial emotion by using Bayesian Network.

Brain-Computer Interface (BCI) is an interface technique between human and computer which can help severely motor-disabled persons to communicate and control their environment. Many researchers use EEG signals for BCI. However, these studies are early stages and it has some problems such as a lack of accurate analysis and measurements with many electrodes. Therefore, we propose an EEG signals measurement method and analysis methods for BCI. Our purpose of this study is to recognize subjectpsilas intention when they move their arms. EEG signals are recorded during the imaginary movement of subjectpsilas arms at electrode positions Fp1, Fp2, C3, and C4. We analyzed ERS (Event-Related Synchronization) and ERD (Event-Related Desynchronization) which are detected when people move their limbs in the mu wave and beta wave. Results of this study showed that ERD occurred in mu waves and ERS occurred in beta waves at C3 during the imaginary movement of right arm. Similarly, ERD occurred in mu waves and ERS occurred in beta waves at C4 during the imaginary movement of left arm. And ERD occurred at Fp1 and Fp2, in both case.

In this paper, we propose to use the AdaBoost algorithm for face detection. AdaBoost is a kind of large margin classifiers and is efficient for on-line learning. In order to adapt the AdaBoost algorithm to fast face detection, use the original AdaBoost algorithm, the original AdaBoost which uses a given features is compared with the boosting along feature dimensions. The comparable results assure the use of the latter, which is faster for classification. The AdaBoost is typically a classification between two classes. This face detection system operates without the aid of initializing stage and realizes automatic face detection system. The overall structure adopts window scanning and image pyramid structure so that various size of face is allowed to be detected. In addition, real-time performance rate can be achieved through constituting strong classifier with extracting a few but efficient weak classifiers by the AdaBoost learning.

Uncertainty of result of context awareness always exists in any context-awareness computing. This falling-off in accuracy of context awareness result is mostly caused by the imperfectness and incompleteness of sensed data, because of this reasons, we must improve the accuracy of context awareness. In this article, we propose a novel approach to model the uncertain context by using ontology and context reasoning method based on Bayesian Network. Our context aware processing is divided into two parts: context modeling and context reasoning. The context modeling is based on ontology for facilitating knowledge reuse and sharing. The ontology facilitates the share and reuse of information over similar domains of not only the logical knowledge but also the uncertain knowledge. Also the ontology can be used to structure learning for Bayesian network. The context reasoning is based on Bayesian Network for probabilistic inference to solve the uncertain reasoning in context-aware processing problem in a flexible and adaptive situation.

Security is an important space, the necessity of intelligent surveillance systems are becoming incredibly important. Most of the security system is an object tracking system based on images which has many weaknesses. This paper proposes the system which complement the vulnerability. Purpose of this paper is loading network camera and tracking module of sound source in mobile robot based on embedded Linux for tracking intruder. The images obtained by network camera and the informations of sound source tracking module are transmitted through wireless LAN. Server grasps the mobile course of the object by extracting objectpsilas motion vector and gives the objectpsilas direction information to a robot by combining the course data and direction data of the sound. We can track the intruder with this method. The robot tracks the intruder according to gets the effective image of an intruder. In compliance with the this paper the system which is embodied protection system step of existing is effective and reliability all with the intelligent surveillance system which uses an image and the sound source in the mobile robot.

In the moving of the mobile robot, the mobile robot acquires a map of its environment while simultaneously localizing itself relative to the map. Simultaneous localization ad mapping (SLAM) problems arise when the robot does not have access to a map of the environment, nor does it know its own pose. In this paper, we study the SLAM of mobile robot in the indoor environment with Digital Magnetic Compass and Ultrasonic Sensors. Digital Magnetic Compass has a strong feature against interference in the indoor environment better than compass which is can easily be disturbed by electromagnetic sources or large ferromagnetic structures. Ultrasonic Sensors are cheap and can give relatively accurate range readings. Autonomous mobile robot is aware of robot's moving direction and position by the restricted data. Also robot must localize as quickly as possible. As application for the SLAM on the autonomous mobile robot system, robot can find the localization and the mapping and can solve the Kid Napping situation for itself. Especially, in the Kid Napping situation, autonomous mobile robot use Ultrasonic sensors and Digital Magnetic Compass(DMC)'s data for moving. When robot receives the similar data by sensors, robot uses Computation Intelligence(CI) for perceiving in the robot's position.

In swarm robot systems, each robot must behaves by itself according to the its states and environments, and if necessary, must cooperates with other robots in order to carry out a given task. Therefore it is essential that each robot has both learning and evolution ability to adapt the dynamic environments. In this paper, reinforcement learning method with SVM based on structural risk minimization and distributed genetic algorithms is proposed for behavior learning and evolution of collective autonomous mobile robots. By distributed genetic algorithm exchanging the chromosome acquired under different environments by communication each robot can improve its behavior ability. Specially, in order to improve the performance of evolution, selective crossover using the characteristic of reinforcement learning that basis of SVM is adopted in this paper.

This paper presents evolvable neural networks based on a developmental model for navigation control of autonomous mobile robots in dynamic operating environments. Bio-inspired mechanisms have been applied to autonomous design of artificial neural networks for solving practical problems. The proposed neural network architecture is grown from an initial developmental model by a set of production rules of the L-system that are represented by the DNA coding. The L-system is based on parallel rewriting mechanism motivated by the growth models of plants. DNA coding gives an effective method of expressing general production rules. Experiments show that the evolvable neural network designed by the production rules of the L-system develops into a controller for mobile robot navigation to avoid collisions with the obstacles.

An important problem of pattern recognition is to extract or select feature set, which is included in the pre-processing stage. In order to extract feature set, principal component analysis has been usually used and SFS (sequential forward selection) and SBS (sequential backward selection) have been used as a feature selection method. This paper applies genetic algorithm which is a popular method for nonlinear optimization problem to the feature selection problem. So, we call it genetic algorithm feature selection (GAFS) and this algorithm is compared to other methods in the performance aspect

Simultaneous localization and map building (SLAM) problem is occurred under the situation that a robot is moving on the unknown place. In this paper, We study the SLAM of the mobile autonomous robot system. We assume that the moving space of robots is the Ideal indoor space. Also, we suggest the data format to build the map. The data format is only used for this robot system. But it may be useful to the other robot systems. Most important thing that we present in this paper is using the electrical compass. We can get the points of the electrical compass. That point represents an original direction of the robot. Precisely, in this research, the robot uses the electrical compass and vision camera to solve the kidnapping. The robot has to deal with the points of the compass as the map building parameters. As an application for the SLAM on the autonomous mobile robot system, this study is aimed to optimize the performance of the robot navigation on the 2-D space with vision camera and electrical compass. So this paper suggests the new data format to solve the SLAM problem. Moreover, how the robot can find its own position in the kidnapping situations for itself. The robot has to be more intelligent and smart. The method is sufficient to the robot applications.

This paper presents and proposes two approaches to speed up for the controller unit. One is a residue number system (RNS), the other is a sorting network. Using an RNS reduces the total bit size of adder and carry-propagation delays. Also, using sorting network makes comparing time of 16 input data sequences within 10 clocks. First, this paper introduces the basic concept of an RNS and a sorting network. Next, we design the controller using the RNS adder and Green model 16-input sorting network. We present experimental results about elevator group control by applying these two approaches. Experimental results show that the RNS can reduce adder size, which calculates an elevator attractive function (EAF) value, to 5-bit and its carry propagation delay within a single residue. The results also demonstrate that a sorting network is able to choose the best elevator to serve against passenger's request by comparing the attractive function value of each elevator with in 4 clocks. It means that our primary goal can be achieved by these two approaches.

Recently, the research of processing sensor information has been performed actively. At this time, we need a context recognition manager system that can aware near circumstances and act adaptively, though the same device is used. If a machine could recognize user's circumstances, it can reduce the failure on working and improve the user's working ability by using the recognized information. In this paper, we propose a system that is adaptive to a user. After the system perform a predefined action by using merged information with circumstance's context and user's emotion, it adapts itself to the change in its surround or user's emotional response.

Signature based intrusion detection system (IDS), having stored rules for detecting intrusions at the library, judges whether new inputs are intrusion or not by matching them with the new inputs. However their policy has two restrictions generally. First, when they couldn't make rules against new intrusions, false negative (FN) errors may are taken place. Second, when they made a lot of rules for maintaining diversification, the amount of resources grows larger proportional to their amount. In this paper, we propose the learning algorithm which can evolve the competent of anomaly detectors having the ability to detect anomalous attacks by genetic algorithm. The anomaly detectors are the population be composed of by following the negative selection procedure of the biological immune system. To show the effectiveness of proposed system, we apply the learning algorithm to the artificial network environment, which is a computer security system.

This paper deals with the creation of planetary gear trains including sensor for torque and rpm and the useful way to measure the torque and RPM of the geared motor. For this we have made the planetary geared reduction motor with Hall sensors and with the monitoring system. The monitoring system displays the sensing data (torque, rpm) and calculated data (power) and also has the network capability with Bluetooth protocol. We show that our solution is much more inexpensive and simple method to measure torque and rpm than before.

Distributed autonomous robotic system (DARS) consists of central processing part, communication part, sensor part, and etc. It is essential for DARS, which is made of independent agents. Bluetooth is developed having the features as low power consumption, standard protocol, small module size and etc. So it is suitable for DARS. But it has some constraints like network structure, distance of communication, and so on. To overcome these constraints, more dynamic network system is needed for DARS. In this paper, we propose the self-organizable bluetooth network base on the routing-tree that is more adaptable for DARS. This network system is composed of the following four network-organization procedures-discovery, organization, maintenance, and re-organization (C. Chevallay et al., 2002). Also this system can operate in more dynamic environment and organize communication network by itself. By the reasons, this self-organizable bluetooth network is suitable for DARS and other dynamic environment.

Recently, robot technique has been developed remarkably. But, they cannot do emotional tasks and the work they can is limited. In this view, it is important for machine to understand human's emotion. Also, emotion recognition is necessary to make an intimate robot. This paper shows simulation results which classify emotions by learning 4 pitch patterns and results from some analyses. The pitch contour includes emotion information. This is why the pitch has been widely used for emotion recognition. However, because the pitch contour is not sufficient for recognizing emotion, we should add other elements. Thus, several analyses are done and the analyzed elements are called acoustic elements for convenience. These elements are the fundamental for more accurate recognition. In addition to this, we analyze the relation between emotion and acoustic elements. The brain is high-dimensional nonlinear dynamical system. So, it is essential to utilize a system that is capable of storing internal states and utilize a system that is capable of storing internal states and implementing complex dynamics. DRNN fits such a system. The simulator is composed of the DRNN (dynamic recurrent neural network), feature extraction.

It enables small antibodies to detect numerous pathogens which the natural immune system maintains the variety of B cells and T cells. The combination of genes being located in the V-region, D-region and J-region of B cell's chromosome can generate tens of thousands of B cells. Also T cell receptor in the T cell has four kinds of receptor protein molecules-a, /spl beta/, ?, d chain-which maintain the variety of T cells using the combination of them. After the T cells being recombined are passed through the strict verification procedure in the thymus, only T cells which survive can detect the intrusion of external pathogens. In this paper, we introduce the maintenance mechanism for the variety of T cells and propose the algorithm which detects the intrusion using detectors composed of those which are maintained after being passed through the negative selection procedures in the thymus. Finally, we prove the efficiency of the algorithm using computer simulation.

Emotion recognition has various methods. Mainly, it can be performed by visual or aural methods. We know that it is possible to recognize people's emotion by only sound data. We use the pitch of speech as a main feature. We define features of four emotions (normal, angry, laugh, surprise) in pitch analysis. Based on this feature pattern, we implement a simulator using VC++. First of all, this simulator is composed of 'generation of individuals', 'recurrent neural net (RNN)', and 'evaluation'. Using the result from the learning part of this simulator, we can get results applied to other speech data (excepting for learning data). In detail, each module uses the following method. First, the generation of individuals part uses (1+100)-ES and (1+1)-ES (that is, random). Thus, we observe the comparison result of both methods. Then, we select the best way. Second, the RNN part is composed of 7-nodes. That is, 1 input node, 2 hidden layer nodes, 4 output nodes. Selection of this structure depends on the characteristics of sequentially inputted speech data. Third, the evaluation part is very important. This part is the cause of the extraction speed and satisfaction degree of result. Then we implement a simulator from the above modules. Applied to other speech data, we observe the result of recognition.

We propose a fuzzy classifier system (FCS) to find a set of fuzzy rules which can carry out the edge detection. The classifier system of Holland (1985) can evaluate the usefulness of rules represented by classifiers with repeated learning. The FCS makes the classifier system be able to carry out the mapping from continuous inputs to outputs. It is the FCS that applies the method of machine learning to the concept of fuzzy logic. The antecedent and consequent of a classifier is same as a fuzzy rule. In the paper, the FCS is the Michigan style. A single fuzzy if-then rule is coded as an individual. The average gray levels which each group of neighbor pixels has are represented in a fuzzy set. Then it is decided whether a pixel is an edge pixel or not by using fuzzy if-then rules. Depending on the average of the gray levels, a number of fuzzy rules can be activated, and each rule makes the output. These outputs are aggregated and defuzzified to take a new gray value of the pixel. To evaluate this edge detection, we compare the new gray level of a pixel with the gray level obtained by other edge detection methods such as Sobel edge detection. This comparison provides a reinforcement signal for the FCS which is reinforcement learning. Also the FCS employs genetic algorithms to make new rules and modify rules when performance of the system needs to be improved.

We present a new type of neural architecture consisting of chaotic neurons and apply it to the prediction of chaotic time series signals. To evolve chaotic neural systems, we use cellular automata whose production rules are evolved based on a DNA coding method. The structure of networks are appropriate for learning nonlinear, chaotic, and nonstationary systems. In order to verify their effectiveness, we apply the evolutionary chaotic neural systems to one-step ahead prediction of Mackey-Glass time series data.

The fuzzy classifier system (FCS) makes the classifier system be able to carry out the mapping from continuous inputs to outputs. The classifier system can evaluate the usefulness of rules represented by classifiers with repeated learning It is the FCS that applies this ability of the machine learning to the concept of fuzzy controller. It is that the antecedent and consequent of classifier is same as a fuzzy rule of the rule base. In this paper, the FCS is the Michigan style and fuzzifies the input values to create the messages. The system stores those messages in the message list and uses the implicit bucket brigade algorithms. Also the FCS employs the genetic algorithms (GAs) to make new rules and modify rules when performance of the system needs to be improved.

In this paper, we propose a method of cooperative control (T-cell modeling) and selection of group behavior strategy (B-cell modeling) based on the immune system in a distributed autonomous robotic system (DARS). The immune system is a living body's self-protection and self-maintenance system. These features can be applied to decision making of optimal swarm behavior in a dynamically changing environment. To apply the immune system to DARS, a robot is regarded as a B-cell, each environmental condition as an antigen, a behavior strategy as an antibody and control parameter as a T-cell respectively. When the environmental condition changes, a robot selects an appropriate behavior strategy, and its behavior strategy is stimulated and suppressed by other robots using communication. Finally, the most stimulated strategy is adopted as the swarm behavior strategy. This control scheme is based on clonal selection and idiotopic network hypothesis. It is used for decision making of the optimal swarm strategy. By T-cell modeling, the adaptation ability of the robot is enhanced in dynamic environments.

An adaptive neural-fuzzy control scheme for intelligent control is proposed. The control system consists of a fuzzy-neural controller (FNC) and model neural network (MNN). In the FNC, the antecedence and consequence of the fuzzy rule are constructed by a clustering method and a multilayer neural network. In the MNN, a multilayer neural network is utilized to identify an unknown controlled plant. The error backpropagation algorithm has been adopted as a learning technique. The effectiveness of the scheme is demonstrated by computer simulations of a cart-pole and a two-d.o.f. robot manipulator.

In this paper, we present the reinforcement learning and distributed genetic algorithm based behavior learning of the distributed autonomous mobile robots. The internal reinforcement signal for the reinforcement learning is generated by fuzzy inference, and dynamic recurrent neural networks are used as action generation module. We adopt the distributed genetic algorithms for the cooperative behavior emergence. We show the validity of the proposed learning and evolution algorithm by computer simulation.

In this paper, we propose a method of cooperative control based on immune system in distributed autonomous robotic system (DARS). Immune system is living body's self-protection and self-maintenance system. Thus these features can be applied to decision making of optimal swarm behavior in dynamically changing environment. For the purpose of applying the immune system to DARS, a robot is regarded as a B lymphocyte (B cell), each environmental condition as an antigen and a behavior strategy as an antibody respectively. The executing process of proposed method is as follows: when the environmental condition changes, a robot selects an appropriate behavior strategy, and its behavior strategy is stimulated and suppressed by other robot using communication. Finally, such stimulated strategy is adopted as a swarm behavior strategy. This control scheme is based on clonal selection and idiotopic network hypothesis, and it is used for decision making of an optimal swarm strategy.

The theoretical foundations of genetic algorithms (GA) are the schema theorem and the building block hypothesis. In the meaning of these foundational concepts, simple genetic algorithms (SGA) allocate more trials to the schemata whose average fitness remains above average. Although SGA does well in many applications as an optimization method, still it does not guarantee the convergence of a global optimum. Therefore as an alternative scheme, there is a growing interest in a co-evolutionary system, where two populations constantly interact and co-evolve in contrast with traditional single population evolutionary algorithms. In this paper, we propose a new design method of an optimal fuzzy logic controller using a co-evolutionary concept. In general, it is very difficult to find optimal fuzzy rules by experience when the input and/or output variables are going to increase. So we propose a co-evolutionary method finding optimal fuzzy rules. Our algorithm is that after constructing two population groups made up of rule base and its schema, by co-evolving these two populations, we find the optimal fuzzy logic controller. By applying the proposed method to a path planning problem of autonomous mobile robots when moving objects exist, we show the validity of the proposed method.

We propose a method of cooperative control (T-cell modeling) and selection of group behavior strategy (B-cell modeling) based on the immune system in distributed autonomous robotic system (DARS). The immune system is a living body's self-protection and self-maintenance system. These features can be applied to decision making of optimal swarm behavior in a dynamically changing environment. For applying the immune system to DARS, a robot is regarded as a B-cell, each environmental condition as an antigen, a behavior strategy as an antibody and control parameter as a T-cell respectively. The executing process of proposed method is as follows. When the environmental condition changes, a robot selects an appropriate behavior strategy and its behavior strategy is stimulated and suppressed by other robots using communication. Finally much stimulated strategy is adopted as a swarm behavior strategy. This control scheme is based on clonal selection and the idiotopic network hypothesis. It is used for decision making of optimal swarm strategy. By T-cell modeling, adaptation ability of robot is enhanced in dynamic environments.

This paper presents a genetic programming based evolutionary strategy for on-line adaptive learnable evolvable hardware. Genetic programming can be a useful control method for evolvable hardware for its unique tree structured chromosome. However it is difficult to represent the tree structured chromosome in hardware, and it is difficult to use the crossover operator in hardware. Therefore, genetic programming is not as popular as genetic algorithms in the evolvable hardware community in spite of its possible strengths. We propose a chromosome representation method and a hardware implementation method that can be helpful for this situation. Our method uses a context switchable identical block structure to implement a genetic tree in evolvable hardware. We compose an evolutionary strategy for evolvable hardware by combining the proposed method with other research results. The proposed method is applied to the autonomous mobile robots cooperation problem to verify its usefulness.

With the rising innovative antigens (such as intruders and viruses) through the Internet, reliable security mechanisms are required to perceptively detect and put them down. However, defense techniques of the current host system over the Internet may not properly analyze Internet antigens, because trends of attacks are unexpectedly shifted. We introduce an Antibody Layer that mediates proper security services based on the biological mechanism to rapidly disclose and remove innovative antigens. The proposed Antibody Layer also employs a new topology called antibody cooperation protocol to support real time security QoS for one host as well as host alliance.

The authors propose a method of constructing neural networks using bio-inspired emergent and evolutionary concepts. This method is an algorithm that is based on the characteristics of biological DNA and the growth of plants. The authors propose a construction method to make a DNA coding method for production rule of L-system. The L-system is based on the so-called parallel rewriting mechanism. The DNA coding method has no limitation in expressing the production rule of L-system. Evolutionary algorithms motivated by Darwinian natural selection are population based search methods, the high performance of which is highly dependent on the representation of solution space. In order to verify the effectiveness of our scheme, we apply it to one step ahead prediction of Mackey-Glass time series and Sun spot data.

We propose a new type of evolution strategies combined with reinforcement learning. We use the change of fitness occurred by mutation to form the reinforcement signals which estimate and control the step length of mutation. With this proposed method, the convergence rate is improved. Also, we use Cauchy distributed mutation to increase the global convergence faculty. Cauchy distributed mutation is more likely to escape from a local minimum or move away from a plateau than Gaussian distributed mutation. After an outline of the history of evolution strategies, we explain the evolution strategies combined with the reinforcement learning, that is reinforcement evolution strategies. Performance of the proposed method is estimated by comparison with conventional evolution strategies on several test problems.