Based on a new concept of Kinder-GuaRdian which aims to realize a support system for kindergarten's staffs and parents by watching over children's activity and their growing process, activity recognition (AR) for children is treated in this paper. In case of children, smaller amount of sensors is better since easy fitting of the sensor device to child's body is strongly preferable. In this study, using one acceleration sensor attached to the upper arm, AR in cases of adults and children were evaluated using several orthodox classifiers and self-organizing map (SOM). Through the several types of validation, effectiveness, issues, and benefits of their methods are discussed.

For intelligent human-machine systems supporting user's operation, prediction of the user behavior and estimation of one's operational intention are required. However, the same high abilities as human being are required for such intelligent machines since human decides own action using advanced complex recognition ability. Therefore, the present authors proposed a Bayesian intention estimator using Self-Organizing Map (SOM). This estimator utilizes a mapping-relation obtained using SOM to find transition of the intentions. In this paper, an improvement of the Bayesian intention estimator is reported by considering the task context. The scenario of whole task is modeled by Petri net, and prediction of belief in Bayesian computation is modified by other probability estimated from the Petri-Net scenario. Applying the presented method to an estimation problem using a remote operation of the radio controlled construction equipments, improvements of the estimator were confirmed; an undetected intention modes were correctly detected, and inadequate identification was corrected with adequate timing.

An estimation algorithm of operational intentions in the machine operation is presented in this paper. State transition relation of intentions was formed using Self-Organizing Map (SOM) from the measured data of the operation and environmental variables with the reference intention sequence. Operational intention was estimated by stochastic computation using a Bayesian particle filter with the trained SOM. The presented algorithm was applied to the remote operational task, and qualitative and quantitative analyses were performed. As a result, it was confirmed that the estimator could classify the types of intentions as similarly as the human analyst discerned. Further, several issues, such as difficulty in preparation of objective normative data, and necessity of consideration of scenario / causality, are discussed.

This paper addresses evaluation methods for social cooperation characteristics in cooperative tasks. Using these we found that the ratio of four typical indexes relate to task performance. Most of conventional human-machine systems assumed to assist only a single operator. In such cases, the system should only pay attention to his/her intention or operation characteristics. If there were multiple operators they also included social behaviours affected from others. Such behaviours could not be regarded by the one-on-one assist systems. Since intelligent assist systems are expected to apply in human society, the system requires work among humans. A challenge of our research is to quantify the human social characteristics, especially when relating to cooperation in the task involving multiple operators. In this paper, human social characteristics are quantified in order to evaluate cooperative performance. For the evaluations, ratio indexes by four kinds of behaviours are focused on. One is observation behaviour for other operator's work, which is typical social behaviour in a cooperative task. Second is the ratio of approaching to other robots for physical supporting. Third, total activity by summation of robot motion distance. Then, an index of visibility of robots each other. Finally, correlation between expansion of the index distribution and a task completion time is analyzed. Furthermore advanced assist using the evaluation is discussed.

Model of a human-machine system to express the process of cognition-intention-action is presented by referring a concept of an attention in Global Workspace Theory. Based on the model, new approach to estimate intentions of an operator who manipulates machines is proposed by utilizing Self-Organizing Map (SOM) and the Bayes filtering. Probabilities of transition of intentions are approximated by SOM from the operational log data, and the intentions are estimated through a Bayesian particle filtering with the trained SOM. The effectiveness was verified by applying to the remote operational task, and potential of the intention estimator was confirmed. Several issues were analyzed, and guideline for further improvement is discussed.

For manipulation of remote mobile robots, adequate scheduling of tasks and selecting of operational commands are required. This paper presents an analysis procedure to make the task switching profile visible by utilizing the Self-Organizing Map (SOM) and new cluster growing method. For practical verification, an experiment system with radio-controlled construction equipments was built, and the proposed analysis procedure was applied to the experimental task. As a result, it was confirmed by correlation analysis that distances among decomposed clusters corresponding to segments of operation strongly relate to performance index of the task.

This paper addresses a quantification method for role sharing in cooperative tasks. By the method, we found that the ratio of three typical indexes relate to task performance. Most of conventional human-machine systems assumed to assist single operator. In such case, the systems should only pay attention his/her operation characteristics. If there were multiple operators, they also include altruistic behaviours during the work. Such behaviours could not be regarded by such one-on-one assist system. Since intelligent assist systems are expected to apply in human society, the systems are required work among humans. A challenge of our research is to quantify such humans' altruistic behaviours, especially what relating cooperation in the task involving multiple participants. In this paper, role sharing characteristics are quantified in order to evaluate the cooperative performance. For this quantification, ratio indexes of three kinds of behaviours is focused on. One is observation behaviours for other operator's work, which is typical altruistic behaviour in a cooperative task. Second is ratio of egoistic behaviour as active conveyance. Then, total activity by motion distance of the robot could be indexes of the role sharing. Finally, correlation between these ratio of indexes and a task performance is analyzed. Furthermore an advanced assist using the evaluation is discussed.

On a remote vehicle manipulation, both a task scheduling and an adequate selecting of operation commands are significant to execute sequential task successfully. Hence, for a future assistive intelligent mechatronics, automatic recognition of the operator's task-switching is required to support him / her timely. In the present research, an analysis procedure to make the task-switching strategy visible is presented by utilizing the Self-Organizing Map (SOM). For practical verification, a cooperation task in a virtual computer space was used, and the proposed analysis procedure was applied to the experimental data. As a result, obvious separation of each operation mode was demonstrated as the operator became skilled. And, it was confirmed that the proposed method was valid for a skill evaluation of remote vehicle manipulation.

For a future advanced machine, the function for supporting a human by considering his or her skill-level is required. In order to implement such intelligent functions in the machine, it is necessary for the system designers to know what skill of the machine operation is. In this paper, discrete movements of hand reaching actions were focused on, and relations between performances of the machine operation and the reaching action were investigated by referring Fitts’ law. A teleoperation experiment system was built using two radio-controlled construction machines, and the training process of operators was analyzed. The results showed that simple and similar types of reaching actions were repeated at the beginning of the trials and the actions were changing into various modes as the operators became skilled.

The In the modern society, human beings interact with computer networks and mechanical systems. This interaction is, unfortunately, asymmetrical at this point of time. Human beings easily understand the computer's way of thinking, because computers and mechanical systems are human-made. On the other hand, computers do not understand human way of thinking, because computers do not have psychological model of human beings. This situation must be improved so that the interaction between humans and computers could be symmetrical. The physical interaction between humans and computers is done through mechatronics. Mechatronics is known as the discipline integrated by mechanical, electrical and information technology and has been used to produce advanced artifacts used in modern society, which has been developed through the interdisciplinary studies of diverse engineering fields. Our modern life is surrounded and enhanced by gadgets and gimmicks of mechatronics products, and those who use the products can get used to improve their operation skill. It is expected to develop mechatronics systems which are capable of adapting themselves to the level of the skill or dexterity of humans who use the systems. We need to found a new mechatronics discipline considering human in the closed loop. This may be named "Human adaptive mechatronics" and is a new discipline of the integration of mechanical, electrical, information technology and human sciences. Its application fields are found in diverse fields from smart human-machine interface from smart human-machine interface for tele-surgery operation to rehabilitation assisting machines.

This paper presents results in the development of a light-weight, easily portable vision-based eye-gaze tracking device for use in the development of human adaptive mechatronic (HAM) devices. In particular, this paper focuses on the development of a two stage pupil centre locator based entirely on integer calculations. The first stage gives an approximate pupil location while the second stage refines the result of the first stage. This two stage approach allows graceful degradation - an approximate 1/sup st/ stage result can be used should 2/sup nd/ stage processing fail. Results from a software based implementation of this algorithm show that our integer based method has similar localization accuracy to floating point ellipse fitting techniques.

In this research, we show two experiments using non-verbal interaction based on 'humatronics' which construct natural interaction between human and systems. Recently, we use various types of systems based on electronics. However, these systems are not useful and imposes burden. The electronics for solving these problems are called "humatronics". Advanced humatronics are mostly needed for natural communication between human and system. Thus, we think nonverbal communication is basic, and important for communication between human. Moreover, We focus on the networked robotics which is fusion of network technology and robot technology. We constructed nonverbal interface by using human natural motion. In the first experiment, we created commands using hand motion which is based on natural motion, such as becking, hissing away and so on. We constructed robot manipulation system using these commands. Furthermore, we compared the user-friendly of our system with present keyboard manipulation system. In the second experiment, in order to manipulate a robot using more natural motion, we proposed 'pointing movement'. Moreover, we constructed networked robots that recognizes human motion spatially. Accordingly, the user will be able to manipulate the robot with more natural motion in larger space.

We propose ontology for an agent communication with human and other agents according to environmental information (situation). Recently, various types of robot and information tool are developed and used. But now, the system is still promoted and developed by human. The system isn't able to build interaction between human. The agent processes information is led by a single robot, and isn't able to utilize relationship with human or other agents. We consider the informational aspects of intimacy recognition technology, and information transport and sharing. Therefore, we focus attention on ontology' technology. Human can communicate with other, by considering the situation and the body language. This is because human has the ontology. In this research, ontology is composed of conceptual fuzzy sets. Conceptual fuzzy sets can combine separate concept one by one, and form whole concept. The agent created the ontology front observed human motion and situation, and communication with human or other agents using ontology. We show the composition of ontology from the interaction between human and agent, and experiment of the communication with robot using ontology.

Presents a method for reducing odometry errors caused by kinematic imperfections in wheeled mobile robots. A new velocity trajectory is proposed that compensates for orientation error due to acceleration constraints on motor controllers. According to this velocity trajectory, the wheel velocity of one out of two driven wheels may be changed by the traveled distance of the mobile robot. It is shown that a wheeled mobile robot can't move along a straight line exactly, even if kinematic imperfections are corrected perfectly, and this phenomenon is attributable to acceleration constraints on motor controllers. We experiment on a wheeled mobile robot with 2 DOF and discuss the results.

We deal with the design of a WMR's fuzzy controller for trajectory tracking. The fuzzy controller uses a genetic algorithm to optimize membership function and reorganize the term set, and fuzzy rule. A similarity measure is taken to change the term set and rule correction is performed based on the changed term set. We propose the modified algorithm to take care of fine tuning. The searching region is gradually cut down while the modified GA is performed.

Real-time obstacle avoidance is essential for the safe operation of mobile robots in a dynamically changing environment. This paper investigates how an mobile robot can respond to unexpected obstacles while following a path planned by a global path planner. The obstacle avoidance problem is formulated using a decision theory to determine an optimal response based on inaccurate sensor data. It is shown that our proposed method avoids properly obstacles through simulation results.

The artificial potential field method is frequently employed for obstacle avoidance both of mobile robots and of manipulator arms. This paper extends previous results for planar problems to the general n-dimensional case. The complexity is reduced by projecting the n-dimensional workspace onto a 2D subspace called operation plane. Furthermore, only the closest obstacles are taken into account when designing the artificial potential field. Still, global convergence is achieved. The methodology is illustrated with several numerical examples.

Summary form only given. Discusses how robotics should essentially be a subject of human behavior. We cannot expect a robot to achieve high performance such as excellent preciseness, high speed, high power. Such requirements are accomplished by specialized machines not by robots. We have to pay more attention to the study of human behavior in our robotics research. This requires us to have wider viewpoints including new disciplines such as ecology and sociology, etc. I propose a concrete strategy to make progress in robotics research.

A key problem in robotics is the estimation of the location and orientation of objects from surface measurement data. This is termed pose estimation. A fundamental task is the pose estimation of known quadratic surfaces from, possibly noisy, data. A solution for this task facilitates pose estimation for more complex objects. Current algorithms frequently converge to local minima of the performance index and/or pay a high computing cost and/or are sensitive to noise, that are unsuited for online applications because of the intensive computer effort required. The goal is to develop a fast and robust algorithm for pose estimation using range data. Here, pose estimation is carried out using algebraic techniques in a two stage optimization procedure involving least squares estimation, or better the method of instrumental variables, and 3/spl times/3 matrix diagonalizations. The procedure leads to zero pose estimation error in the noise free finite data case, and in the case of infinite data with additive white noise.

Proposes a position estimation method and a path generation between current position and goal point for mobile robots. Dead reckoning has been commonly used for position estimation. However this method has inherent problems because it also accumulates estimation errors. In this paper, the authors propose a new method to increase the accuracy of estimated positions using a matching method which is applied with a least squared scheme. This approach uses features, such as corner points and edges of the object in the task environment instead of land-marks. Also, the authors discuss how to establish an environment database and propose a genetic algorithm in order to find an optimal path. It is shown that: it is possible to estimate the position of mobile robot precisely, errors are not cumulated, and the path generation method is very fast.

Intelligent cooperative manipulation is proposed as a new paradigm for human-machine interaction and cooperation. Aiming at intelligent assistance of human operators the information and power flow between operator-system-task environment is carefully analyzed using a virtual reality to simulate the task environment. Central issue of this virtual reality is the dynamic force simulator (DFS) for feedback force calculation proposed in this paper. The DFS simulates object dynamics, contact model and friction characteristics of the human hand interacting with objects in the virtual reality. An example shows the efficiency of the proposed realization scheme of the DFS.>

The author deals with the present status and future perspective of power electronics and motion control. This discussion is particularly relevant today because some electrical engineers believe that motion control is a part of power electronics. However, one belief is that power electronics and motion control are now distinct fields-despite having common technologies and shared roots.>

The approach presented is this paper is based on two ideas: (1) on the planning level an electrostatic potential field in the configuration space of the robot is used. The generalized force curves are attractive to the goal point avoiding obstacles without local minima; (2) on the control level the system follows these force curves using sliding mode control. The system is attracted to the swiching manifold using sliding mode and therewith the systkm is robust to disturbances and parameter variations. the system control variables so that sliding mode occurs in the vicinity of the force field curves and the system is attracted to them. This means that the system trajectories are identical to the force field curves. Sliding mode assures robust control performance despite unknown disturbances and parameter variations.

The approach described consists in constructing potential fields in configuration space like an electrostatic one with changes distributed in such a way that the generalized force curves are attracted to the goal point and avoid obstacles. The system follows such force curves by using sliding mode. The strategy of robot path obstacle avoidance control with sliding mode is presented and discussed in the two-dimensional space.>

VSS controller is suited for robotic arms where the robust performances in the presence of parametric variations and disturbances are most important. The practical design of such robust VSS controller is discussed in this paper. The significant issue in the design of VSS is to avoid the chattering caused by the switched input. This problem is solved by introducing the continuous control input instead of switched input. The practical design of such control is obtained by using "Balance Condition". The "Balance Condition" is derived from the careful consideration on the bandwidth of plant dynamics and VSS controller, and gives many benefits in applications of VSS controller to actual systems. This paper shows the validity of balance condition in the robotic arm control. Simulations and experimental results using a position servo system (one degree of freedom robotic arm) are discussed from the practical point of view of robust control. In order to investigate multi-joint cases, simulations with a two-linkage robot arm are provided.

The high-gain effect of sliding mode control based on VSS (Variable Structure System) suppresses the uncertainties due to parametric variations, external disturbances and variable payloads. The resulting system is completely robust whereas the obtained control law is simple and easy to be applied to on-line computer control. In this paper, the sliding mode controller for MIMO (Multi-Input Multi-Output) robot arm is realized using "Estimated Inertia Matrix" which has inevitably small uncertainties dependent on arm structure and inaccuracies of computation. This technique contributes to practical design of sliding mode for MIMO system compared with the well-known "Hierarchy Method". In addition, simple nonlinear compensation and proper continuous function (instead of relay type component) with "Integral Mode" are implemented to diminish chattering caused by switching input. The validity of this control is confirmed in simulations and experiments where the system shows the robust performances in spite of the existing nonlinear interactions and unknown parametric changes.