In this paper, we propose a novel method for a humanoid robot to recover its standing posture from different fallen postures. In the proposed method, the robot estimates its posture using the ground-contacting information, which is sensed by the joint angle information and pressure sensors attached to the whole body. Then, the robot searches a motion sequence based on the estimated posture to stand up. The robot preserves every successful motion sequence into the tree structure to be reused in the next time. To verify the proposed algorithm, we conducted evaluation experiments by employing a real robot. The results show that the proposed method enables the robot to successfully search the motion sequence for standing up from the fallen posture.

The papers in this special section focus on the concept of laughter computing. Laughter is considered a significant feature of human-human communication. Laughter is characterized by a complex behavior that includes major modules: auditory, facial expressions, body movements, and postural attitudes, and physiological signals. The goal of this special section is to gather recent achievements in laughter computing in order to trigger new research directions in this area.

We developed a four-arm four-crawler advanced disaster response robot called OCTOPUS. Disaster response robots are expected to be capable of both mobility, e.g., entering narrow spaces over very rough unstable ground, and workability, e.g., conducting complex debris-demolition work. However, conventional disaster response robots are specialized in either mobility or workability. Moreover, strategies to independently enhance the capability of crawlers for mobility and arms for workability will increase the robot size and weight. To balance environmental applicability with the mobility and workability, OCTOPUS is equipped with a mutual complementary strategy between its arms and crawlers. The four arms conduct complex tasks while ensuring stabilization when climbing steps. The four crawlers translate rough terrain while avoiding toppling over when conducting demolition work. OCTOPUS is hydraulic driven and teleoperated by two operators. To evaluate the performance of OCTOPUS, we conducted preliminary experiments involving climbing high steps and removing attached objects by using the four arms. The results showed that OCTOPUS completed the two tasks by adequately coordinating its four arms and four crawlers and improvement in operability needs.

We have developed an interactive 4-D visualization system that employed the principal vanishing points operation as a method to control the movement of the eye-point and the change in the viewing direction in 4-D space. Different from conventional 4-D visualization and interaction techniques, the system can provide intuitive observation of 4-D space and objects by projecting them onto 3D space in real time from various positions and directions in 4-D space. Our next challenge is to examine whether humans are able to develop a spatial perception of 4-D space and objects through 4-D experiences provided by the system. In this paper, as the first step toward our aim, we assessed whether participants were able to get intuitive spatial understanding of 4-D objects. In the evaluation experiment, firstly, the participants learned a structure of a hypercube. Then, we evaluated their spatial perception developed in the learning period by tasks of controlling the 4-D eye-point and reconstructing the hypercube from a set of its 3-D projection drawings. The results indicated evidence for that humans were able to get 4-D spatial perception by operating the system.

Much attention has recently been given to a printing method because they are easily designable, have a low cost, and can be mass produced. Numerous electronic devices are fabricated using printing methods because of these advantages. In paper mechatronics, attempts have been made to fabricate robots by printing on paper substrates. The robots are given structures through self-folding and functions using printed actuators. We developed a new system and device to fabricate more sophisticated printed robots. First, we successfully fabricated complex self-folding structures by applying an automatic cutting. Second, a rapidly created and low-voltage electrothermal actuator was developed using an inkjet printed circuit. Finally, a printed robot was fabricated by combining two techniques and two types of paper; a structure design paper and a circuit design paper. Gripper and conveyor robots were fabricated, and their functions were verified. These works demonstrate the possibility of paper mechatronics for rapid and low-cost prototyping as well as of printed robots.

Recently, there has been strong interest in printed robots and paper electronics. Printing is an adequate manufacturing method for the mass production; robots manufactured by printing can be low cost, rapidly creatable and easy to design. Paper is one of smart materials with high water absorbency and strength; it is easily mass-produced at low cost. Many inexpensive and flexible devices were created in paper electronics. Therefore, we propose paper mechatronics that merges printed robots and paper electronics. We used a commercial ink jet printer with water-based ink for self-folding of paper; the paper was folded automatically along the printed line to make the robot body. The paper robot was equipped with a printed electrothermal actuator that consists of epoxy resin covered by conductive ink. The epoxy resin was heated by Joule heat of the conductive ink under a driving voltage of 8 V. The self-folding of experimental paper robot took 15 minutes to construct the body and the robot demonstrated a locomotion at 6.5 mm per step.

In this paper, we report a capsule gel robot that has a novel motion ability. The robot is produced by wrapping a chemical reaction system in a gel membrane and demonstrates a rotary motion by moving the center of gravity. We employed an oil-water system wrapped in a capsule where an oil droplet moves autonomously in water phase. We introduced an alginate gelation technique to make spherical hollow gels, in order to fabricate a membrane of capsule gel robot. The produced capsule gel robot showed a rotary motion in water.

This paper presents a novel approach for stabilizing a two wheeled mobile robot while traversing down a stair terrain. During flight, the drive wheels of the mobile robot are used as control input to generate reaction torque to alter the pre-impact landing angle, ensuring a successful landing. A state dependent feedback control scheme is used for continuous step traversing while keeping the robot upright. Relationships between travel velocity, height of step and landing angle were investigated. The proposed approach was simulated using physics simulation engine and a physical prototype was constructed to confirm simulation findings. Results were compared with conventional two wheeled robot control scheme and effectiveness of proposed approach was confirmed.

Generally, the bottom-up learning approaches, such as neural-networks, are implemented to obtain the optimal controller of target task for mechanical system. However, they must face a problem including huge number of trials, which require much time and give stress against the hardware. To avoid such issues, a simulator is often built and performed with a learning method. However, there are also problems that how simulator is constructed and how accurate it performs. In this study, we are considering a construction of simulator directly from the actual robot. Afterward a constructed simulator is used for learning target task and the obtained optimal controller is applied to the actual robot. In this work, we picked up a five-linked manipulator robot, and made it track a ball as a task. Construction of a simulator is performed by neural-networks with back-propagation method, and the optimal controller is obtained by reinforcement learning method. Both processes are implemented without using the actual robot after the data sampling, therefore, load against the hardware gets sufficiently smaller, and the objective controller can be obtained faster than using only actual one. And we consider that our proposed method can be a basic and versatile learning strategy to obtain the optimal controller of mechanical systems.

This paper describes a novel approach that can be used to determine the focal object that a robot is looking at and introduces the unique robotic eye, AttractiveEye. Humans have the visual input mechanism but do not have the output mechanism concerning the visual signal except gestures. AttractiveEye can communicate information regarding the color of an object by changing its color to match that of the object. Thus, from the color of this robot, we can determine the focal object that it is looking at. We carry out a human-robot interaction experiment to show that human-robot interaction can be made easier by using AttractiveEye.

In our previous study, we first succeeded in construction of a novel-types self-oscillting molecular robot. However, the driving environment of the molecular robot was firmly restricted in strong acid conditions. This is because the molecular robots drive induced by the Belousov-Zhabotinsky (BZ) reaction, which is well known for exhibiting temporal and spatiotemporal oscillating phenomena. The overall process of the BZ reaction is the oxidation of an organic substrate, such as malonic acid (MA) or citric acid, by an oxidizing agent (bromate ion) in the presence of a strong acid and a metal catalyst. In this study, in order to drive the novel molecular robot under the physiological condition, we conducted the modification of the molecular structure of the self-oscillating polymer chain. In order to cause the self-oscillation under the biological condition, we synthesized a built-in system where the BZ substrates other than organic acid were incorporated into the molecular robot itself. As a result, the novel molecular robot drives under the biological condition. We believe that the development of the novel molecular robot lead to construction of the novel biomimetic soft robots and actuators, and may inspire novel nonlinear experimental and theoretical considerations.

In this proceeding, we introduce peristaltic gel actuators as a chemical robot. The polymer gels prepared here have a cyclic reaction network like metabolic process in itself. With a cyclic reaction, the polymer gel swells-shrinks autonomously. The periodic self-oscillating motion of the gel is produced by the dissipating chemical energy of the oscillatory Belouzov-Zhabotinsky (BZ) reaction. We have succeeded in convey the object automatically by utilizing the synthetic polymer gel. This experimental fact represents the great possibility of the chemical robot.

In our previous studies, we succeeded in construction self-oscillating gel actuator such as a self-walking gel and a dynamic motion of a gel actuator. However, the speed of the swelling-deswelling self-oscillation is too slow in order to realize practical chemical robot. In this study, we succeeded in the wide-range control of the period for the swelling-deswelling self-oscillation of a novel polymer gel by selection of the initial concentration of the Belouzov-Zhabotinsky (BZ) substrates and the temperature. The novel polymer gel was composed of a non-thermoresponsive and biocompatible poly-vinylpyrrolidone (PVP) main chain covalently-bonded to the ruthenium catalyst for the BZ reaction. Moreover, we clarified the influence of the initial concentration of the BZ substrates and the temperature on the period and the self-oscillating behavior. By optimizing the initial concentration of the BZ substrates, we cause the swelling-deswelling self-oscillation in 0.5 Hz. The maximum frequency (0.5 Hz) of the novel gel was 20 times higher than that of the conventional-type self-oscillating gel. Moreover, we showed that the displacement of the self-oscillation for the gel has a trade-off relationship against the frequency of the self-oscillation.

epsiv-filter can reduce most kinds of noise from a single-channel noisy signal with preserving the signal that varies drastically such as a speech signal. The filter design is simple and it can effectively reduce noise. It can reduce not only stationary noise but also nonstationary noise. However, it has some parameters and we need to set them appropriately based on empirical control. So far, there are few studies to evaluate the appropriateness of the parameter setting of epsiv-filter. In this paper, we employ correlation coefficient of the filter output and the difference between the input and the filter output as the evaluation function of the parameter setting. We also show the algorithm to set the optimal parameter of epsiv-filter. We conducted the experiments to compare the value of the correlation coefficient and the mean absolute error when we change the value of epsiv. The experimental results show the applicability of our criterion in parameter setting of epsiv-filter.

This paper describes an algorithm for internal noise reduction from dependent mixtures of microphones and a piezoelectric device under blind condition. The features of our approach are summarized as follows: (1) The purposed method handles internal noise reduction of the acoustical array combining microphones and a piezoelectric device in the time domain. (2) It can reduce the internal noise from each microphone even when the number of the sounds and noise is larger than that of the acoustical sensors. (3) It does not utilize the statistical independence of the sounds. The sounds may be not only the statistically independent but may also be statistically dependent. (4) It handles the partially disjoint sounds in the time domain. The sounds may be overlap in the frequency domain. (5) It does not require to estimate the intervals when sounds are disjoint. We formulate the problem and show the methodology of the proposed method. Experimental results are also demonstrated to confirm the effectiveness of the proposed method.

Self replication is a kind of self-assembly in which a desired structure is obtained automatically from simple parts. We propose self-replication not by controlling the parts directly, but by controlling their environment, namely passive self-replication, which makes self-replicating structures by imitating the semi-conservative replication of DNA. The proposed self-replication strategy is available for multiple types of parts by introducing selective surface structures on parts and two different bonding forces.

This paper presents a highly accurate and efficient method for crack detection using percolation-based image processing. The detection of cracks in concrete surfaces during the maintenance and diagnosis of concrete structures is important to ensure the safety of these structures. Recently, the image-based crack detection method has attracted considerable attention due to its low cost and objectivity. However, there are several problems in the practical application of image processing for crack detection since real concrete surface images have noises such as concrete blebs, stains, and shadings of several sizes. In order to resolve these problems, our proposed method focuses on the number of pixels in a crack and the connectivity of the pixels. Our method employs a percolation model for crack detection in order to consider the features of the cracks. Through experiments using real concrete surface images, we demonstrate the accuracy and efficiency of our method.

Passive-dynamic walkers are mechanical devices that walk down a slope without motors or controllers. In this paper we present our research in two distinctive parts. First, a design improvement on the classical four-legged passive-dynamic walking machine and second, an investigation on the timing of different stages in the human walking cycle and comparison of the results with the results obtained from our walker. We built a four-legged passive-dynamic walking machine with its inner and outer legs connected rigidly two by two, making it equivalent to a biped machine from a dynamics point of view. It features a new mechanism for an active release of the locked extended knee instead of using the more common knee-locking via suction cups mechanism. We conducted a series of experiments with this modified walker in which we counted the steps it made while walking down an incline. We compared the results with a previous design of the knee-locking mechanism that used permanent magnets. The improved model made an increased number of successful walks down the slope from which we concluded that the new active release mechanism is more reliable and easy to use and set up. For the walking cycle investigation, we put visual markers on a person walking casually down the same slope that we used for the walker experiments. We took a video of that person and measured the times between different events in one walking cycle. We calculated the average time intervals and made a comparison between the test subject and our walker within one cycle. We feel the timings derived from this investigation can be used to make the gait of our walker more human-like.

This paper presents a tactile interaction system for human-machine interaction and an active sensing technique to realize object surface sensing for human-machine interaction tasks. The tactile sensor unit implemented on the robot hand consists of three force-sensitive resistors arranged triangularly. The unit not only can detect the reactive force from the object but also determine the orientation of the object surface. Our research goal is to realize an effective robot controller for the 3D object handling. As a possible application, we also introduce an effective human-robot cooperation to move an object together. This application is realized by a control of a hand pose to keep the direction of the hand normal to the object surface in 3-D, which is often necessary when pushing an object. We also show the experimental results using the proposed sensor unit.

Stimuli-responsive polymers and gels have been applied to biomimetic actuators or artificial muscles. Biomimetic robots using such materials can move like a living creature. Especially, electroactive polymers are said to be promising materials as advanced soft actuators. Its mechanical motion is controlled by external stimuli such as a change in solvent composition, pH, temperature and electric field etc. On the other hand, many living organisms generate an autonomous motion without external driving stimuli. In this paper, we report a novel biomimetic gel actuator that can bend and stretch with worm-like motion spontaneously without external driving stimuli. The gel actuator converts the chemical energy of oscillating reaction, i.e., the Belouzov-Zhabotinsky (BZ) reaction into the kinetic energy inside the gel. Although the gel is completely composed of synthetic polymer, it shows autonomous motion as if it is alive. To cause anisotropic contraction with curvature changes, the gel strip with gradient structure was prepared. Furthermore, the gel can walk with repeated bending and stretching motion by itself like a looper by coupling with ratchet mechanism. The "self-walking" gel actuator will create a new framework of biomimetic chemical robot.

This paper describes a novel approach of autonomous navigation for a mobile robot based on a passive RFID that will be used in human living environment. Conventional approaches using dead reckoning or landmark are influenced by disturbances such as the light condition and obstacles, while the proposed method can obtain environmental information more reliably and robustly with the RFID system. The proposed method estimates not only the position but also its current orientation without other sensors by using the detected sequential IC tag information. We examined that the robot reaches to the goal utilizing only the position information but not the orientation information.

In this study we analyze the performance of an ensemble model composed of several perceptrons that can effectively deal with nonlinear classification problems. Although each member of the ensemble can only deal with linear classification problems, through a competitive training mechanism, the ensemble is able to automatically allocate a part of the learning space that is linearly separable to each member, thus decomposing non-linear classification problems into several more manageable linear problems. In this paper, we gave the performance analysis of the ensemble with regard to some benchmark problems.

Image transmission delay is a serious problem for a remote operator when he/she controls a robot according to the images transmitted by the robot through a narrow band width channel or from extremely distant place. We propose a simple method to improve the remote control operability by introducing the estimated image sequence that is generated from the delayed images. The robot is equipped with an omni-directional camera to make the processing easy. The image shown to the operator is estimated from the transmitted image according to the robot's motion and the delay time. In the case of forward movement, the received image is magnified and displayed, while it is shifted in case of rotational movement of the robot. As the operator can see the estimated present image without time delay, he/she does not miss the timing of the controls. This paper describes the proposed image estimation method with some experimental results which demonstrate the validity of our method

Information processing technology can be classified into three historical phases. The first phase is the physical information processing to treat physical data from the real world. This technology is often called "signal processing". As the target is mainly physical signal such as sound, brightness and force, the base of the processing is the law of nature. The reality of the system is examined by the physical explanationability of the input-output relations and the causality is important. The second phase is the logical information processing to deal with knowledge and rule. "Artificial intelligence" is included in this phase where symbols and languages are processed according to the rule. The most important is consistency and provability here. We are entering the third phase of information processing that is "KANSEI" information processing to treat human emotion. "KANSEI" is a Japanese word expressing some subjective concept including "sensibility", "feelings", "intuitiveness" and "mood". The term "human factor" has a close relation with KANSEI. However, in the KANSEI information processing, we intend to approach the human emotional world more positively by applying computer technology to the affective information processing and human communication including art, musical performance and entertainment

In the past some decades, the advanced information technology has proposed a variety of new ways of music creation and performance to make people free from the physical limitations of instruments by providing different abilities on music creation. This paper describes some interface devices and software tools that have been developed in our group

In recent years, we can easily access an enormous amount of digital video stream in standardized video format such as MPEG etc. However, it is not so easy to find a desired video stream from a video database in a reasonably short time. Some efficient searching methods in terms of computational cost are definitely required for video stream retrieval. In this paper, we propose a novel method for video stream retrieval using a video stream as search-key. The objective is to find a part of the video stream stored in the database that is similar to the given video stream key. This method extracts some characteristics of frame transitions in video streams, and utilizes the characteristics for the similarity matching. Some experimental results are also shown to verify the efficiency of the proposed method.

In this paper, we introduce a novel method of relevance learning by a multi-layer perceptron. The relevance learning is regarded as learning from the relationship among two or more outputs of the network. The learning network architecture is based on a simple multi-layer perceptron with a modified back-propagation learning algorithm. Unlike the conventional multi-layer perceptron that learns from a set of an input feature vector and the target output, the proposed network can obtain a nonlinear mapping between a set of two or more vector inputs and the desired relevance. For instance, the desired relevance represents the dissimilarity among given objects. We show the performance of the proposed network with some experiments with four artificially generated data set. We then discuss the theoretical and mathematical background underlying the network learning with some related works. We evaluate the obtained arrangement of objects in comparison with the result of principle component analysis (PCA) and multidimensional scaling method (MDS). This work also contributes to the measurement of human subjective evaluation for multidimensional perceptual scaling. Some experimental results on the low-dimensional representation of color hue data set and emotional facial images were presented.

In This work we propose a modified GA that assigns a unique mutation rate to each gene based on the contribution of the respective gene's contribution to the fitness of the individual. Although the proposed model is not "parameter free", through a number of experiments, we show that the parameters for this model are significantly insensitive to the landscape of the problems compared with the mutation rate in conventional GA, implying that this model could deal effectively with a wide range of problems the requirement to set the mutation rate empirically.

In this study we introduce an ensemble of neural networks, in which each member is a linear perceptron. Our main objective is to build an ensemble of neural networks that can automatically and effectively divide the problem space and assign a subspace to each member. By assigning only a portion of the problem space, we expect that the learning difficulty for each member can be reduced, thus leading to better classification ability. To investigate the effectiveness of the proposed method in dividing the problem space, in this paper we deal with ensemble consists only of linear perceptrons, each with an additional output neuron that indicates the confidence level of the output.

A substantial robotic interface is proposed for collaborative work between humans and machines in a multimodal musical environment. The robotic interface is regarded as a "moving instrument" that displays the reactive motion on a stage while producing sound and music by embedded stereo speakers according to the context of the performance. In this paper, we introduce four musical platforms utilizing robotic technology and information technology in different circumstances. These are effective designing environments for artists such as musicians, composers, and choreographers, not only for music creation but also for media coordination including motion and visual effects. The architecture, called the MIDI network, enables them to control the robot movement as well as to compose music. Each of the developed robotic systems works as a sort of reflector to create an acoustic and visual space with multimodality. The proposed approach to equip musical instruments with an autonomous mobile ability promises a new type of computer music performance.

In this research we propose a model of autonomous robot that is able to survive in real world outdoor environment without human intervention. For an autonomous robot, energy is a very important factor, so it is only natural that the ability of the robot to accumulate energy should be used as evaluation of the robot's survival strategy. The robot in this study is equipped with a solar panel, and a model of neural network ensemble. The solar panel enables the robot to autonomously accumulate energy in the outdoor environment, while the neural network ensemble enables the robot to acquire strategies that are needed in order to survive in the environment. The accumulated energy is considered as the feedback from the environment to measure the efficiency of the strategy exerted by the robot, and used as learning guidance for the robot to survive.

In this paper, we propose a model of a neural network ensemble that can be trained with a supervisor having two kinds of input-output functions where the occurrence probability of each function is not even. This condition can be likened to a learning condition, in which the learning data are hampered by noise. In this case, the neural network has the impression that the learning supervisor (object) has a probabilistic behavior in which the supervisor generates correct learning data most of the time but occasionally generates erroneous ones. The objective is to train the neural network to approximate the greatest distributed input-output relation, which can be considered to be the principal nature of the supervisor, so that we can obtain a neural network that is able, to some extent, to suppress the ill effect of erroneous data encountered during the learning process.

In this paper, we propose a computer aided diagnosis system that facilitates the drafting of treatment plans, the simulation of orthognathic surgery and the prediction of postoperative features. The three-dimensional (3D) display of the oral and maxillofacial region is very efficient for dentists in understanding a patient symptom and drafting an appropriate treatment plan. To achieve this, we propose a construction method for a 3D head model, which consists of soft and hard tissue. Utilizing feature points extracted from ortho-directional X-ray images and facial images, a standard head model can be modified to adapt an individual head shape. Using this head model, we can predict surgical modification of soft and hard tissue.

The genetic algorithm (GA) has been successfully introduced to solve various optimization problems. One of the characteristics of the GA is that, once it has converged, most of its population members are copies of the best individual, causing the GA to lose population diversity. This characteristic is a setback when we consider non-stationary problems in which the fitness functions vary with time. In this paper, we propose a migrational GA that stores past environmental solutions and retrieves them rapidly when that environment is re-activated, through probabilistic operation.

We have previously (1999) proposed a model of a neural network ensemble composed of a number of multi layer perceptrons (MLP). The ensemble is trained so that each member has a unique expertise. It is also provided with a mechanism to automatically select the most relevant member with respect to the given environment, enabling the ensemble to adapt effectively in changing environment. In this research we trained the ensemble with a noisy training data set, which is a training set that contains a particular percentage of contradictionary (false) data. Based on the members' expertise the ensemble has the ability to distinguish contradictionary data and treat such a kind of data set as one unique environment that differs from the clean environment formed by correct data. In the training process the ensemble will automatically select one of its member to be trained in the clean environment and switch to another member whenever a contradictionary data is given, resulting that one of the ensemble member will be successfully adapting the clean environment.

The paper presents a new method for a trend diagnosis system using neural networks. Most dynamical systems are not easy to analyze and faults are difficult to detect because the observed parameters do not directly express the state of the system. We need to measure the temporal tendencies of the parameters, which isn't easy for testing machines or humans. The effectiveness of the trend fault diagnosis system using recurrent neural networks is examined for an air-conditioning system. The network was trained with faulty and correct data sequences obtained from system simulation. The experimental fault detection results using actual data proved that the proposed method is effective for performing trend diagnosis of dynamic systems.

The air-conditioning systems of buildings have been diversified in recent years, and the complexity of the systems has increased. At the same time, stability in the system and low running cost are demanded. To solve these problems, various research projects have been done. The development of the energy load prediction systems and the fault detection and diagnosis systems have received great attention. The authors propose a real time fault diagnosis system for air conditioning units (the heating unit, the cooling unit, the air intake unit, and the air-recycling unit) using a recurrent type neural network.

Presents statistical analyses of the search-space landscape of knapsack problems in due consideration of stochastic optimization. It is known from existing works that travelling salesman problems have a rugged landscape. We deal with 1000 knapsack problems where the values and the weights of the objects are arranged randomly. By introducing proper estimate values and topology, it is revealed that the landscape of these knapsack problems is a rugged landscape similar to that of the travelling salesman problems. It is assumed that the rugged landscape is a combination of a global valley-like structure and a local noise-like structure. We propose a new algorithm to estimate the optimum point by introducing the least mean square method to fit the global structure at some points selected randomly in the search space. The method does not contradict "no free lunch" theorems because of the availing features of the landscape. It is forecast that not only knapsack problems but also many practical problems have a structure which is characterized with the same measure. These results are useful in order to compose more effective optimization methods without a trial-and-error process.

In this paper we propose a model of neural network ensemble composed of a number of multilayer perceptrons (MLP), each with a unique expertise. Using temperature control the most appropriate ensemble member will be automatically activated for a given environment while the irrelevant members will be inhibited. The proposed temperature control will enable the neural network ensemble to work efficiently in multiple environments.

The Humanoid Project which the authors are engaged in is introduced with the recent results on the image processing technology for robot. Topics include a multimedia sensing system to track a speaking person using both image and sound processing and a model-based vision system to recognize the self position in the known environment.

The purpose of our work is to realize an agent for intelligent interaction between humans and robots, embedding a computational model of artificial emotions with learning and self-adaptation features. The gesture of a human can cause the robot to change its "emotional state", which is exhibited by means of integrated visual media, environmental lights, music, and changes in robot's style of movement and behavior. We adopt our emotional agent architecture for developing the agent. The system is a kind of "emotional activator", which plays the role of stimulating the human creativity. In our environment, humans do not need to wear any special on-body sensor. We use the small robot on wheels, equipped with on-board ultrasound sensors. Furthermore, We develop a camera-based sensor system to increase its input capabilities.

Though many motion interfaces have been proposed, most of them are based on the handle manipulation such as a mouse and a joystick. With these interfaces, users must concentrate on handling. We propose a new interface "plate" where a user can show any direction simply by shifting his/her body. The center of gravity is measured by four strain-gage elements bonded under the plate. As a primary evaluation of this interface, we constructed an experimental system to connect the plate to a mouse interface and examined the behaviors of novice users. As the second step of the development we implemented the plate on an omnidirectional vehicle as an active motion system.

In this paper, we propose a new method for curve detection based on the inverse Hough Transform. The key idea of this method is to make the voting process on the image space instead of that on the parameter space in the conventional method, then convert the local peak detection problem in the parameter space into a parameter optimization problem. This leads to substantial savings, not only in storage requirements but also in the amount of calculation required. The experimental results and qualitative analysis showed that in comparison with the conventional Hough Transform methods, the new method has advantages of high speed, small storage arbitrary parameter range and high parameter resolution.>