Moving object detection from video sequences is an important problem in robot vision and In-vehicle camera. However it is difficult when the camera itself is moving. In this paper, we propose a method to detect a moving object by applying anomaly detection method using Cubic Higher-order Local Auto-Correlation (CHLAC). Our method is applicable to various camera motion without an assumption of itself moving model, and has an advantage in robustness to noise and calculating amount. In experiments, we apply this method to detection of human motion when the camera is moving and verify the effectiveness.

View full abstract

Commonly moving objects in the footage are generally detected with computing the difference between the background image and input images. However, it has the problem that the shadow of moving objects are also considered as moving objects. This paper proposes a shadow detection method using correlation analysis of the area regarded as a moving object and its evaluation using trajectries of moving objects.

View full abstract

In order to develop intruder detection system or provide user-focused services to walking persons, we proposed the method to estimate their destination from trajectory of walking persons with fixed camera using Hidden Marcov Model. To evaluate availability of our method, we calculated discrimination ratio for elapsed time, and confirmed it reaches to 80% after 2 seconds of observation. However, it is not enough to evaluate availability, because the ratio of approach to destination is not reflected in this index. In this paper, we introduced the ratio of approach to destination as another index of availability, and present the calculation result of both of them. Additionally, we discuss about how to use them properly depending on the situation.

View full abstract

Humanoid robots are expected to work in human environment due to their similarity to the shape of human, and are required to achieve localization and navigation autonomously. In this paper, indoor navigation is realized based on a view-based approach using a camera mounted on the head. In the case of humanoid robot, image blur and swing due to walking is a crucial issue for image matching during localization. The quantitative effect of walking on the image is firstly investigated utilizing a motion capture system. Then a method to generate a stable view-sequence is proposed based on detecting optical flows. The navigation function with the proposed method was implemented on a humanoid robot HRP-2, and the effectiveness was confirmed by indoor walking experiments.

View full abstract

This paper presents a method to analyze a relationship between postural adjustment and ocular motion. Vestibular nuclei integrate the signals from VOR and OKR learning pathways. And also, there is a nerve pathway from the vestibular nuclei to spinal motor neurons. Reflex ocular motion has an intimate involvement in involuntary movement such as bipedal locomotion. Reflex ocular motion and involuntary movement are simply modeled based on the concept of compound control. Self-organized map is used for learning algorithm. The features of VOR and OKR are mapped on extraocular muscle motor neurons and the features of signals integrated between a semicircular canal and a retina are mapped on the motor neurons for involuntary movement. We hope our proposal method helps to reveal mechanisms for emergence of functional cooperation in the brain.

View full abstract

In this paper, we propose a human detecting method for mobile robots that uses RFID and stereo vision. For detecting a people who stands still with a stillness state we extract outline domain of human in depth image using human's aspect ratio. On the other hand for detecting a dynamic people we use a method of a difference between adjacent frames. The paper details the architecture of the proposed method.

View full abstract

Many objects are discriminated using the information from the sense of sight, even if it is in the sunlight or under moonlight. Robots need advanced sensor like a human eye to adapt to the environment and chase a target and so on. The target of our research is realized vision system using the vision sensor with fast processing, low-power, small system and include wireless communication system. In this paper, prototype of vision module was developed using a silicon retina and proposal mechanism has 2 guide rails and 2DOF like universal joint. In the moving tests, utility of the mechanism that can move smoothly for 2DOF is made sure.

View full abstract

A mixed analog-digital integrated vision sensor was designed to detect an approaching object in real time. To respond selectively to looming stimuli, the sensor employed an algorithm inspired by a visual nervous system in the locust, which can avoid a collision robustly by use of visual information. An electronic circuit model was designed to mimic the architecture of the locust nervous system. Computer simulations showed that the model provided appropriate responses for collision avoidance. We implemented the model with a compact hardware system consisting of a silicon retina and FPGA circuits and the system was confirmed to be available in real-world scenes.

View full abstract

This paper proposes a method for model-based motion tracking by multiple cameras using 2D-Distance-Field. The 2D-Distance-Field represents errors between the captured images and projected images of 3D model on the image plane. Using the 2D-Distance-Field, which is constructed rapidly using the Fast Marching Method, 3D motion of human is estimated robustly in real time. This paper shows some experimental results including simultaneous motion tracking of multiple people.

View full abstract

In recent years, intelligent robots that work automatically have become required. Conventional robots have been used in specific environments, but these new robots are going to be used in universal environments that are not constant. For that purpose, it is necessary for a robot to be able to recognize the outside environments. In this study, we attach two CCD cameras to a robot to provide information on the outside environment. It is necessary to detect the position and the angle of an unknown inclined plane for perpendicular contact. In general, the three dimensional position of the target is obtained by stereo analysis, but there is no technique to detect the unknown inclined plane. So we propose the new angle detection with vision. As a result, the practicability of this method is verified.

View full abstract

Tins paper describes a new recon.flgurabie processor architecture for high frame rate visual processing. This architecture employs a 2-D mesh processing element (PE) array in which the PEs can he configured to operate as SIMD arrays or operation-pipeline trees depending on image processing algorithms so that maximum on-chip memory consumption is reduced, To achieve high on-chip memory utilization, the architecture features that. the instruction register in each PE is mapped in its local memory space and that tire ALU network and the local memory network can be configured independently. Experimental results show that the proposed architecture effectively utilizes both of the SIMD and operation pipeline modes.

View full abstract

This paper proposes a method for identifying an acceleration sensor from many moving objects in a camera image. The canriera and the acceleration sensor are connected with a network and they are not synchronized. To estimate the time lag and evaluate the correlation, NCC (Normalized Cross-Correlation) is utilized. Since the coord.iizate system of the acceleration sensor is not known, NCC of the iiorrns of the acceleration for the cainiera and the sensor are computed. The experimental results show that the proposed method successfully identifies the acceleration sensor from two moving objects.

View full abstract

The PAS (Physical Agent System) is the remotely controlled autonomous robot. The PAS consists of an autonomousmobile robot subsystem (PAR: Physical Agent Robot), operation subsystem. In this paper, I propose smart vision platform and the hardware compensation method of camera image, using FPGA for PAS. The smart vision platform consist of FPGA, multiple cellular phone CMOS camera and embedded Linux.

View full abstract

We have been developing Assistive MObile Robot System (AMOS), which helps a daily life of physically handicapped persons. The aim of AMOS is to replace only simple tasks that to transport daily use objects in the indoor environment semi autonomously. If the operator chooses the object to be transported in the environment, then AMOS approaches to the objects, pick up it, and finally relocate it in the designed location. To allow the user to operate AMOS easily, we have been developing vision based user interface which accepts the gesture of the operator as a command. As a first step, only simple gesture clipping is implemented and evaluated.

View full abstract

In this paper, we propose the lip shape recognition system for voice recognition for small sized robot. In noisy environment, it sometimes fails to recognize human voice. So, we developed the lip shape recognition system for subsystem of voice recognition aimed to improve the rate of recognition. We achieved the lip shape recognition using small vision system and got the good experimental results about that.

View full abstract

Robots that share workspaces with human are expected in the field of services in the home and office, medicine, and so on. In this research, the neck movement was used for as instructions method for the robot. Furthermore, "KANSEI" transfer function is introduced to convert human instructions into acceptable movement for human emotions. Here, it is thought that the parameter of "KANSEI" transfer function is related to the moving direction and relative position between a human and the robot. The parameter values of the "KANSEI" transfer function is adjusted based on the moving direction and the relative position between human and the robot.

View full abstract

Handshake is an embodied interaction with contact. In our previous research, a handshake approaching motion without a voice greeting was analyzed, and a motion model was proposed based on the analysis. However, humans usually shake hands in a voice greeting. Therefore, in this paper, we develop a handshake approaching motion model in a voice greeting. In this model, handshake approaching motion is generated from trajectory of human hand using the second order lag element and dead time element. The effectiveness of the model is demonstrated by sensory evaluations using a handshake robot system.

View full abstract

Today, many robot competitions, i.e. robocon, are held. This paper introduce one of such competition named Intelligent Robot Contest. The main feature of this competition is that all robots must be autonomous, i.e., human cannot control the robot, while many of robocons allow manual control. Moreover, tasks of robot is not simple, for example, robot must move around field, search objects, handle and recognize them, and carry them to goal, by itself. Therefore, higher technique are required to competitors. The following content shows rules of the competition and management of it.

View full abstract

This paper describes the development of RM-eye a image processing board for small size robots. It can detect a center of gravity and size of three colors, and it send a motion command for typical robot controllers automatically. Therefore nobody need an acknowledgement of color detect algorithm and programming languages. It is expected that it contributes an increase of participants of robot competition and researches.

View full abstract

RoboCup is an international joint project to promote Al, robotics, and related field. It is an attempt to foster Al and Intelligent robotics research by providing standard problems where a wide variety of technologies can be integrated and examined. For example, in RoboCup middle size league, the robots should recognized object colors (ball is orage, and goals are blue and yellow, etc.). And, in 2006, RoboCup middle size league decide to the color is affected by weather during the game. In this paper, we present the color recognition algorithm for RoboCup. This algorithm is used for color constancy algorithm of using modular network SOM. And, we present color recognition experiment in RoboCup environment.

View full abstract

According as our society again , assist machines begin to needed for persons of advanced age. We research and develop a fitness apparatus according to requirement of elderly people. Proposed apparatus which is consisted of a rowing machine and Functional Electrical Stimulation (FES) can exercise every part of muscle for the persons of advanced age. The rowing mechanism was developed to train rowers. The rowing exercise can train legs and upper effectively. Move over FES can assist the exercise of the leg by using surface electricity stimulus. It confined that the investigated apparatus trained the muscles of the person of advanced age effectively and could cancel exercise shortage.

View full abstract

The functional electrical stimulation (FES) is a method of treatment with which the electric stimulus is given to the stripe and the nervous system of blotting out paralyzed by the stroke and the spinal injury, etc. , and the movement is rebuilt. Externally powered orthosis supplements the muscle force generated by the electric stimulus, and the power assistance orthosis which stands up and can walk is developed in this research.

View full abstract

In the recovery of the arm function trouble such as by the apoplexy, repeating therapeutic exercise is effective. Then, we induced the movement by using the robot, and developed the movement support by which the change in muscular power and the range of motion etc. were able also to do the quantitative evaluation in addition so that an accurate therapeutic exercise was continued. It proposes the developed outline of the system and the therapeutic exercise program of the arm function by the system.

View full abstract

The surgical navigation robot is used in order to position the instrument correctly to the target disease. The development of a MRI (Magnetic Resonance Imaging) compatible navigation robot is also recently started so as to utilize more rich information of the MR imaging. In this research, we discuss about highly precise control of the navigation robot. This robot consists of ultrasonic motors (USM) and nonmagnetic materials. We operate the robot by using the phase difference control circuit which we developed as the driver of the USM and show the control performance by experimental results. In addition, we show the result of the MRI compatibility tests which we inspected using 0.2T open MRI.

View full abstract

In this paper we deal with a simultaneous optimal design problem of dynamic parameters and the trajectory for robot manipulators. The dynamic characteristic of robot manipulators is uniquely determined by base parameters, which are functions of physical parameters, e.g., the mass the moment of inertia and damping, etc.. We propose a new simultaneous optimal design method of base parameters and the trajectory using a numerical design algorithm. The optimal base parameters can be obtained by proposed method for the resultant trajectory.

View full abstract

We report an outline of "Kosen Robocon" and present a summary of our two robots, caterpillar robot and jump-roping robot. Characteristics of our robot are humanlike dynamic jump and simple-two-wheel system. Additionally, we introduce our approach to education for engineer by robot contest. We demonstrated our two robots in a regional event to learn various experiences. Our robots were improved by the experience and try-and-error in the event. In the robot contest 2006, our robot won the tohoku regional cup. Successful experience changes the students' mind. Demonstration of robot in the events leads to regional contributions and rising in presentation skill.

View full abstract

This paper addresses a motion generation mechanism making use of Bayesian Network-based internal model for Human- Robot-Interaction. Rapid progress of robot technologies will bring a symbiosis society with robot to us in the near future. In such situation, since it is expected that novice users will use the robots, development of easier methods to teach motions to robots will be required. In this research, we propose the internal model based on Bayesian- networks that generates the robot motions, and updating methods of conditional probability tables through interaction with humans. We apply this model to an interaction robot developed in our laboratory to verify effectiveness by experiment.

View full abstract

We develop a prototype of arm wrestling robot system for force display communication analysis and support. The system comprises a virtual arm wrestling system for the analysis by synthesis of force display communication characteristics by using anthropomorphic agent and force display system, and an arm wrestling interaction robot system with visual servo for understanding of human emotional change and health care support. The concept of the system is also proposed.

View full abstract

Many robots use the automatic control with image processing recently. However, teiching materials of image processing is too difficult for children to use and too expensive to set the circumstance They are eagerly waiting for developing a cheap and easy teaching material. We made a soccer robot with image processing. The robot is composed of OJ2 and DEB101A+104A. OJ2 is an autonomous robot for playing soccer. DEB101A+104A is an image processing board. The robot uses OJ2 as drive unit and DEB101A+104A as processing unit. We inspected that is it possible playing soccer only with CMOS camera with the robot. As a result, the robot can recognize a color ball and track it.

View full abstract

When a soccer robot decides its position planning for the team strategy, the information of its self location in the field is necessary. This paper describes an observer design for a localization of the soccer robot using measurement of the direction toward a landmark based on an omni-directional camera image. Possibility of the improvement in the observed variables using the distance for the landmark is also discussed.

View full abstract

We propose a novel approach for acquisition and development behaviors through observation of behaviors in multi-agent environment. Observed behaviors of others gives fruitful hints to find a new situation, a new behavior for the situation, necessary information for the behavior acquisition. RoboCup scenario gives us a good test-bed multi-agent environment where a learner can observe behaviors of others during practices or games. It is more realistic, practical, and efficient to take advantages of observation of skilled players than to discover new skills and necessary information only though the interaction of robot and environment. The robot automatically detects state variables and a goal of the behavior through the observation based on mutual information. Reinforcement learning method is applied to acquire the discovered behavior suited to the robot. Experiments under RoboCup Middle Size scenario shows the validity of the proposed method.

View full abstract

The existing reinforcement learning approaches have been suffering from the curse of dimension problem when they are applied to multiagent dynamic environments. One of the typical examples is a case of RoboCup competitions since other agents and their behaviors easily cause state and action space explosion. The keys for learning to acquire cooperative/competitive behaviors in such an environment are as follows:●a two-layer hierarchical system with multi learning modules is adopted to reduce the size of the sensor and action spaces. ●to what extent the other agent task has been achieved is estimated by observation and used as a state value in the top layer state space to accelerate the cooperative/competitive behavior learning. This paper presents a method of modular learning in a multiagent environment, by which the learning agent can acquire cooperative behaviors with its team mates and competitive ones against its opponents.

View full abstract

This study shows a position control method of a six-wheel leg robot. A purpose of this study is to move a robot on the surface of the piping, and to inspect. The system of the robot is composes of main body, control substrate, control PC. This robot has mechanism that can move in all directions using six wheels consisting of differential gear unit. Also this study shows geometrical relation between the six wheel's direction and the rotation center. Furthermore, this study shows the experimental result of turn motion using steers front wheel and a rear wheel.

View full abstract

In industrial plants, straight-line horizontal pipe is typically arranged along with vertical pipe, curved pipe, valves, and flanges. So, it is hoped that the robot can automatically inspect along the pipe. In this paper, we propose the structure of a pipe-surface inspection robot with omnidirectional mobile mechanism. The robot consists of the lower unit and the upper unit. The lower unit has six leg-wheel units that consist of the drive part, the lift part and a magnetic wheel. The upper unit has the differential gear unit and the drive joint that enable a steering, a swing and a bending motion.

View full abstract

In recently, study on skiing turn is researched from various viewpoints. And the mechanism of the skiing turn using the skiing robot is clarified now. However, those results are not concretely reflected in alpine skiing. Therefore, to facilitate theoretical consideration, the passive type skiing robot is developed. The influence on the skiing turn such as the position of gravitational center and the shape of skiing are examined by using this robot. In this paper, it reports on their experimental results.

View full abstract

"Students voluntary study system" of Akita prefectural university was adopted for "University education support program with the trait" of the Ministry of Education, Culture, Sports, Science and Technology in 2003 fiscal year. In our university, it has worked on "Students voluntary study system" since opening the university. This system is that students can decide the subject of study voluntarily, organize the group of study, and the university funds a study and supports students. In the robot study program of our university, the effective degree of this system is large. This time, it introduces the robot study program in "Student voluntary study system".

View full abstract

The game "Rock, Paper, Scissors" is a very popular game since it is easy to play and can be enjoyed by people of all ages. So we produced a robot that can play the game to amuse people. The robot and a person play the game facing each other. We developed an image processing system that is capable of recognizing the shape of a human hand. We then developed a system that can determine the outcome of the game by comparing the hand shape of the human player with that of the robot. We also designed the robot to express feelings through its voice, body actions and facial expressions that varied depending on the outcome of the game.

View full abstract

A long-term experiment of robot therapy for elderly people has been conducted at a health service facility for the aged since Aug. 2003. Three therapeutic seal robots, Paro, were introduced there. This paper describes the results of the experiment. Face scales that consist of illustrations of person's faces were used to evaluate person's moods. As the results, feelings of the elderly people were improved by interaction with the seal robots over 3 years.

View full abstract

This paper describes a method of generating 3D indoor environment model by a mobile robot with laser range finder. Generally, there are more complicated objects in indoor environment than those in outdoor environment, and it is an issue how we model them. First, we obtain 3D range data by moving the laser range finder which is put on an autonomous mobile robot verticaly. Next, we rearrange the data in order to make it easier to generate planes from the rearranged data. Then we generate planes from data. Finally, we make textures which are suitable for planes from the images of cameras, and map them on the planes.

View full abstract

Accurate positioning of a mobile robot and accurate and reliable measurements of surroundings are indispensable for an automated construction of an environment map using a mobile robot. However, the accuracy of conventional positioning methods such as dead reckoning or the use of several landmarks is not high in an unknown and rough environment. To overcome this problem, we have proposed a method named "Cooperative Positioning System (CPS)" , which enables to localize robots with high accuracy even in an unknown and rough environment. This paper proposes the combined system of the CPS and a laser range finder to construct highly reliable environment map. The fifth CPS model named CPS-V and experimental results in a large scaled and unknown environment are introduced.

View full abstract

We propose a new scan matching algorithm that makes use of the simple polygonal structure of environment. Our method extracts line segments from a current scan and matches them with line segments in a previous scan, that is, tracks the line segments in the sequence of scan. Our method simultaneously computes the pose shift (Δx ,Δy,Δθ) based on the correspondence between line segments in the two successive scans and the transformation such that the current scan is mapped optimally to the previous scan. By the use of line segments in the scan, it is possible to detect invariants for position and orientation estimation in the environment. First of all, we explain the algorithm of our method, and then report the experimental result in order o show the effectiveness of our method.

View full abstract

This paper presents a method of map building using interactive GUI for mobile robot. The reason we proposed this method is that it is difficult to generate the accurate map for mobile robot to move safely although many kinds of sensors are used to gather the information of environment. In proposed system, the operator can modify map using GUI by comparing LRF sensor with video stream from web camera.

View full abstract

This paper reports the design of truss structure that considers assembly by a teleoperated dual-arm space robot, analysis of assembly work procedure, and truss parts assembly work experiment that is basic work of assembly, for implementation of the truss structure assembly work assumed as work on the orbit. The assembly work experiment was done by taking particular note of adding a partial automation function to teleoperation by the operator, and generation of three point connected operation of truss structure simultaneously.

View full abstract

In assembly task, when the center line of shaft and hole of mechanical parts are different, assembly fails and parts touch each other. In this paper, we propose the method to search hole position using impedance controlled manipulator. Using this method, assembly robot can move the parts to collect position and try to assemble again.

View full abstract

Force control of a robot arm tip against a moving object's side-surface is achieved by sensor fusion of vision and force. A block is employed as a moving object. Position information of the block's corner is obtained by grabbing an image at frame rate of 33 ms by a CCD camera and processing it. Information on the contact force between the tip of a robot arm and the block's side-surface is obtained by a force sensor. The robot performs accurate profiling tasks to the moving block's surface in real time by using both vision and force information. ART-Linux was used as a real-time operating system.

View full abstract

In this paper, a modeling method to handle furniture is proposed. In real environment, there is a lot of furniture such as drawer, cabinet and so on. If mobile robots can handle these furniture autonomously, it is expected that multiple daily tasks, for example, storing a small object in a drawer, can be performed by the robots. We propose direct teaching based modeling method which can easily give not only how to handle the furniture but also an appearance and 3D shape of it simultaneously. The main topic is about the method of appearance and shape acquisition using external sensors.

View full abstract

We have proposed Robot Assisted Rehabilitation (RAR) in which stimulative impression of robots produces effective rehabilitation. In RAR, mental and social effects are expected in addition to physical effect. With these three effects in mind we have done RAR in the nursing home for the aged and the rehabilitation ward. In this paper, we report trial practices using commercially available entertainment robots in remote operation mode.

View full abstract

Robotic pet was introduced into kindergarten and nursery school class work for 4-6 aged children in order to investigate the capability of the robotic pet as an intellectual instructor and play mate for early childhood educational nursing support. In this work, various kinds of robotic pet, "AIBO" (SONY) , "PARO" (Intelligent System), "ifbot" (Business Design Laboratory) were used for the first step of this study in order to evaluate ability, fascination to child. Child's interest, which is important to derive positive learning attitude toward class activity, was evaluated by counting of the number of raising their hands as the affirmative response of question about robot. Videotaping by DV-camcorder is also applied to find any problems during session. Most of children were much interested in robot, but some problems of using a robot in a classroom were found. It was also observed that they played very kindly with robots.

View full abstract

In this paper, we propose a position estimation method for Multi-Robot system using local communication and communication range. In order to estimate position, connection data among robots is gathered in the host computer. The position among the robots on host computer is renewed using the spring model, until there are no contradiction between connection data and robot location on host computer. Using the computer simulation, this method can estimate the position of robots and we show that the usefulness in the method.

View full abstract