We decide and execute our action from many types of environmental information in our daily lives even if we are not conscious of being guided. Galvanic Vestibular Stimulation (GVS) is known as a method of giving influence on sense of equilibrium. The vestibular system is stimulated by a weak current through an electrode placed on the mas-toid behind the ear. In this paper, we had a motion guidance experiment of walking by using GVS to evaluate the effect. Next, we experimented on the behavioral decision process in human walking phase so that the GVS might influence it more effectively. We make an experiment to judge the avoidance direction by watching the masking movie of oncoming person's body part. From this result, the avoidance judgment is possible because the traveling direction can be controlled by the leg when the leg is lifting and landing. It was clarified that oncoming person's walking phase is important in the behavioral decision in collision avoidance. So we propose the behavioral decision model based at the walking phase.

View full abstract

Recently, various methods that analysed physical movement have been proposed. Kawato has argued that internal model with closed loop between feedback control and feedforward control is useful for sinuous movement as a model for motor control. In this paper, we consider internal model as a function model identified from observed data, and propose a model to extract the characteristic of human movement with singular value decomposition from the time-series data of various sensors. We call the knowledge acquired by the internal model embodied knowledge. In particular, we categorized gesture motions by two kinds of models with the indexes of similarity and estimation using left singular vectors. In addition, the ambulation movement is distinguished by hyperplane of three-dimension constructed by singular value. Finally we discussed characteristic and significance of the movement analysis using singular value decomposition proposed in this paper.

View full abstract

Segmentation plays an important role in behavior analysis. Singular Spectrum Transformation (SST) can extract change-points of human's motion caused not only by changes of muscle forces, but also by those of external forces, and it is effective to transform human's motion data into motion change-points sequences. However, it is difficult to analyze change-points sequences to extract motion pattern changes directly from extracted sequences. In this paper, dissimilarity between two arrays of change-points is defined and a method to visualize the similarity of multiple arrays of change-points is described. Moreover, a method to extract the differences of change-points between two classes of change-points sequences based on sensitivity analysis is proposed. The proposed method was applied to analyze walking motion of different walking speed or different identities. The results prove the effectiveness of the proposed method.

View full abstract

Aiming to design a control method that can drive human-like robots as smoothly and precisely as demanded, we investigated the correlation between muscle activities and hand forces to unravel the mechanism of the human's skillful hand-movement at producing force while manipulating objects, and applied that into the robot control. Considering the agonist-antagonistic muscle ratio, we observed a high correlation estimation of the hand force with the muscle ratio, and derived a role division of muscle synergies in generating the hand force. These results hopefully give an optimal but simple way to control robots so as to make the robot's movement as skillful as human's. To verify this idea, we proposed “synergy based control” method using the characteristics of human's muscle synergies observed, and tested it on a pneumatic-driven arm-robot whose structure assumes the human arm's.

View full abstract

Cello spiccato is known to be very hard to acquire. In this paper, we first identify the forced vibration as an adequate physical model to describe spiccato. We take two typical examples of the forced vibration; swinging and dribbling basketball. Then, from analogies of these two motions, we extract forcing timing and shock absorbing way to give the force as two important physical conditions to realize the forced vibration. Then we discover the ways how to satisfy these two conditions through meta cognitive experiments as well as discussions between two authors, an amateur cello player and a professional. Finally we clarify relationship among technical components to realize spiccato to get better explanation based on the inference diagram.

View full abstract

Brain-Computer Interfaces (BCIs) are promising technologies that control computers by thoughts, and they are used to restore control and communication for severely paralyzed people such as those with amyotrophic lateral sclerosis (ALS). Moreover, BCIs can be appealied to healthy people as well. The P300 speller is one of the BCI applications, which uses the P300: an feature of electroencephalogram (EEG), and it allows users to select letters just by thoughts. The P300 speller presents the user with a matrix containing letters and numbers. The user attends to a character to be communicated and the rows and columns of the matrix are briefly intensified randomly. The intensification of the attended character is served as a rare event in an oddball sequence and it elicits a P300 response. The P300 speller detects the character with the elicited P300 response. However, due to the low signal-to-noise ratio of the P300, signal averaging is often performed, which improves the spelling accuracy but degrades the spelling speed. Predictive text, implemented in the most mobile phones, enables users to spell with fewer key presses; thus, it could improve the spelling speed of the P300 speller. This paper implements a predictive text to the P300 speller, and examines how much the spelling speed for Japanese text is improved. In result, the input time was reduced 21% on an average, 33% at the maximum in the experiments using EEG and th e usability of the proposed system was confirmed.

View full abstract

In this research, we propose a measurement method for the effec t of human relaxation to construct the relaxation sound generation system. At first we perform the electroencephalogram measurement and extract a specific frequency band (θ, α, and β wave) concerning with the human awakening state and catch a tendency of the relaxation by analyzing each content ratio. In addition, we estimate the relaxation degree of the subject in the state hearing sound generated by Interactive Chaotic Amusement System (ICAS) developed in this laboratory to confirm the efficiency of the proposed method. ICAS is the sound generation system to decide the musical note, duration and volume using network of chaotic elements which is able to control synchronousness, and users can generate free sounds by ICAS. As for the result of experiment, we confirmed that the result of the questionnaire evaluation obtained when the sound of ICAS was presented to the subject and the result of the relaxation degree by the proposed method show the same tendency.

View full abstract

Embodiment has been discussed until now from various points of view. For example, the embodiment includes not only simple physical properties but also the individual properties obtained by experience and learning according to the discussion by Merleau=Ponty. Furthermore, the concepts and methodologies to expand the embodiment of a person have been discussed in the research direction of Kukanchi. In this paper, we discuss the multi-robot formation composed of multiple robots from the viewpoint of embodiment. The monitoring by multiple distributed robots is sometimes more effective than that by one robot with high performance. It is necessary for each robot to be aware of its position relative to other robots and to constitute a formation so that a multi-robot performs the monitoring and seeking more effectively as a whole system. Since their facing environment changes every moment, the suitable shape of multi-robot formation also should change. In this paper, we propose a method for constituting the formation of a multi-robot according to the dynamically changing environment. First, we applied a method of multi-objective behavior coordination for integrating behavior outputs from the fuzzy control for collision avoidance and target tracing. Furthermore, we apply a spring model to calculate the suitable target position of each robot for the formation behavior. We discuss the proposed method from the concepts of loose coupling and tight coupling. The tight coupling is realized by the spring model while the loose coupling is realized by the individual decision making based on connection and disconnection with other robots. Next, we propose a method for changing the shape for the formation behavior based on the update of connection relationship and natural length of springs according to the change of environmental conditions. The proposed method is applied to the formation change from triangular formation to column formation according to the change of road width. Furthermore, the proposed method is applied to the monitoring in unknown environments. Finally, we discuss the effectiveness of the proposed method from the viewpoint of embodiment through several simulation results.

View full abstract

In this paper, we propose a decision making method based on reinforcement learning and emotion learning (DRE) for inducing social behaviors of agents. Emotion of animals has an important role in their social interactions. We attempt to incorporate emotion into decision making of agents. To make a social decision making, the DRE combines a decision based on intrinsic fear emotion with a strategic decision obtained by reinforcement learning. In simulation experiments, the effectiveness of the DRE is verified concerning the emergence of social behaviors and the adaptability to an environmental change through an unmoving target search problem.

View full abstract

Nowadays, the aging society is one of big problems in Japan, but the number of caregivers and therapists for elderly people is not enough. Recently, we can get various types of information services easily with the development of information technology, robot technology and network technology. Especially the introduction of robots instead of caregivers and therapists for elderly people is one of the possible solutions. For such a system, human behavior recognition is one of important techniques for a robot to interact with elderly people. Most of previous methods for human behavior recognition are based on off-line statistic approaches for modeling human behaviors. In this paper, we discuss an on-line learning method for human behavior recognition using sensor networks. The developed sysem is composed of sensor network system, robot system, human interface system, and database management server. First of all, we apply a spiking neural network as a learning architecture based on the time series of measured data. Furthermore, we propose a hierarchical learning structure using the spiking neural network for classification of the spatiotemporal patterns. Finally, we discuss the effectiveness of the proposed method through experimental results of human behavior recognition in a living room.

View full abstract

In the conventional cooperative games, it is assumed that the payoff of each coalition is known. However, in the real world problems, there may exist situations in which some coalitional values are unknown. In this paper, we investigate the theory of cooperative games under incomplete information of coalitional values. Assuming the superadditivity, we propose some concepts related to cooperative games under incomplete information. We investigate the solution concepts in a special case when only values of the grand coalition and singleton coalitions are known and show that there exists a focal point solution which is commonly suggested in many points of view.

View full abstract

Today, many people have got the lifestyle-related disease and metabolic syndrome,so they have improved their consciousness for health. It is often said that medicine and one's daily diet are equally important for a healthy body. So, in this study, we propose a method to recommend healthy cooking recipes for meals. The recipes are recommended through the following processes. At first, some recipe candidates are selected based on the constraint on calorie intake, considering the user's schedule. After that, linear programming is applied to the selected candidates with the constraints of carbohydrate, fat, protein, salt, with the objective function of increasing the amount of vegetable intake in the candidates. There are many cases where recommended recipes are not used because of the change of schedule or cooking ingredients. Our proposed method can flexibly deal with such situations.

View full abstract

In conventional studies, it is considered that differences of actuators are not important for realizing intelligent behavior, so, usually conventional robots employ motors as actuators. On the other hand, animals have muscles, and recently, it is reported that physical properties of the muscles like viscosity and elasticity play an important role in realizing intelligent behavior. In this paper, we consider that learning time (time required for adapting themselves to the environment) works as selection pressure, and actuators like muscles are ac quired in evolution. To discuss this hypothesis, we employ a 2-link manipulator, and evolve the manipulator in simulation. The tasks of the manipulator are to avoid an obstacle and to catch a ball. Fitness of the manipulator is calculated from learning time. Simulations have been conducted, and as a result, manipulators that have actuators with adequate viscosity and elasticity have been obtained. By analyzing the result, we have found that the viscosity and the elasticity have reduced learning time.

View full abstract