並木 明夫

In this paper, we propose a robot manipulation system controlled by a stereo high-speed vision system. The vision system recognizes targets by using CAMshift. CAMshift is an tracking algorithm based on the pattern of the color histogram of a target, and it is robust against disturbance of image noise. In the experiment, the system recognized the positions of both a hand of a human operator and a stick grasped by the hand, and estimates the position and orientation of the stick. The robot arm was controlled so as to follow the motion of the stick. The motion of the arm was quick and responsive. The proposed system is useful for human-robot interaction systems.

Humans can perform fast and skillful manipulations using various parts of the body by effectively utilizing the dynamics of the targets. Visual sensation is the most important human sense used for such manipulations. It is obviously preferable that robot can perform various tasks using same end effectors like a human hand and using visual information. Juggling is one such example involving skillful and dynamic manipulations, and visual information is essential for it to be successful. Previously, there have been several studies about robotic juggling. However, none of these studies have considered cases in which a human-like multifingered hand-arm is used for the robotic juggling. The purpose of this study is to achieve two-ball juggling using our robotic hand-arm, which has three general purpose fingers, and stereo vision. The trajectory of the robotic hand-arm is generated based on the ball's estimated dropping position and moment, and the robot catches the ball. The juggling motion is achieved by repeating this cycle.

In recent years, various robot hands and arms have been developed for achieving dexterous manipulation tasks. However there are no robots that can not only move quickly but also handle tools dexterously. Kendama motion is one example of dynamic manipulation and skillful handling. There are several previous researches about robotic kendama. However, in these researches robots have a different configuration with humans, or can not use its robotic hand effectively. The purpose of this study is to achieve kendama motion with a hand-arm by using visual feedback. We apply impedance control to robot hand in order to grasp kendama tool softly. The trajectory of the hand-arm is generated based on optimization. Image processing is executed at 900fps by high-speed vision system and GPU. As a result, the robot has succeeded in kendama motion.

In this paper, we try to achieve origami tasks by multi-fingered hands. Paper is thin and deformable and it crumples easily. If there is a crease in the paper, its stiffness increases in the parallel direction, and its stiffness decreases in the normal direction. Because of these characteristics, paper handling is very difficult. In this study, we analyze human origami motion in 'squash fold', and we propose 4 types of motion of our multi-fingered hands. These are suitable for the structure of our robot hands and it reduces the difficulty of the origami task. We have verified our proposed method by experiments,and the results are shown.

This paper is about a teaching system for a multi-fingered robot hand using kinetic information of a manipulated object with vision feedback. The proposal of this system is to operate a multi-fingered robot hand with high DOF intuitively and effectively. The system calculates a appropriate trajectory according to the kinetic information of the manipulated object under twist rolling contact constraint. This paper describes a revised algorithm of the trajectory generation and the results of experiments using the developed teaching system. The future works to be improved are discussed.

Humans perform skillful and fast manipulations by utilizing the dynamics of the targets effectively. Visual sensation is the most important human sense for such manipulation. It is obviously preferable that robot can perform various tasks using same end effectors like a human hand and using visual information. Juggling is one example of skillful and dynamic manipulation of human, and visual information is essential to juggling. There are several previous researches about robotic juggling. However, there are no researches about robotic juggling in which a human-like multifingered hand-arm is used. The purpose of this study is to achieve two ball juggling using our robotic hand-arm, which has three general purpose fingers, and stereo vision system. The Center of Pressure (CoP) tactile sensors are attached to the fingertips of both thumb and we can measure the grasping condition by using the sensor outputs.

One of the robots to interact with humans is the air hockey robots. The robots require to respond instantly to a high-speed puck. The problems of the present system are control tracking performance in terms of accuracy and generating control input which is out of the range. The goal of this study is to solve the problems. To achieve this, we use Nonlinear Model Predictive Control, NMPC, with C/GMRES method. In this paper, first, the outline of NMPC with C/GMRES method is explained. Secondly, the problem setting of the system is explained. Then, we show the data of experiments and verify the effectiveness of the system.

This paper proposes a robot which can pour liquid from bottle. The robot consists of a 7-axis robot arm and 4-axis robot hand, a high-speed vision, and the robot is controlled based on visual information at the rate of 500Hz. The robot system has the abilities to pour liquid smoothly and exactly, is independent of target. A system designer can easily adjust for target by setting several specified parameters. In this paper, first, vision processing to get liquid surface information. Secondly, pour motion design for pour exactly is proposed. This motion plan consists of 3 step which are extracted from human motion. Finally, the data of experiments are shown and the effectiveness of the propose method is verified.

In the paper, a new master-slave manipulation system is proposed. The system consists of the master system called telexistence FST and a dual-hand-arm robot. The master system measures the pose of a operator by Flexible Sensor Tube (FST) which consists of many bending joints and rotational joints. The weight of the system is light and its movable area is large. Thus, good operability is achieved. The 3D image captured with the cameras of the slave robot is sent to the Head Mount Display (HMD) of the master system. The structure of the slave robot is similar to the upper half of human bodies. It has two 7-axis arms and two multi-fingered hands. Each multi-fingered hand has 5 fingers and 10 axes. Pressure sensors and proximity sensors are set on the surfaces of the fingertips and the forearms. They improve the operability of the system. All joint actuators are controlled through EtherCAT network in realtime. The slave robot sits on an electric wheelchair, and it can move anywhere by pushing its lever. Finally, experimental results are shown, and the validity of the proposed algorithm is verified.

In this research, we focus on a manipulation of a sheet-like elastic object using a high-speed robot hand. Here, we manipulate a card as one of the sheet-like elastic object. In particular, we aim at a card flicking using a fingertip high-speed vibration of the high-speed robot hand. First, we discuss the card grasping at an initial state of the card manipulation based on the geometry condition of the card and the kinematics of the robot hand, and we propose a strategy of the card flicking. Then, we perform an analysis of the card flicking. And, we obtain the condition to flick the card using the slip between the card and the finger of the robot hand. Finally, we show experimental results of the card flicking using the fingertip high-speed vibration of the robot hand.

This paper is about a teaching system for multi-fingered robot hands using kinetic information of a manipulated object with feedback control. The purpose of this system is to operate multi-fingered robot hands with high DOF intuitively and effectively. The system calculates the appropriate trajectory according to the kinetic information of the manipulated object under the twist rolling contact constraint. For more stable manipulation, visual and force feedback control using kinetic information of the real manipulated object is appended to this teaching system. This paper describes the method of feedback control and the future works are discussed.

This paper proposes a robot which can play air hockey game with a human. The robot consists of a 4-axis robot arm and a high-speed vision, and the robot is controlled based on visual information at the rate of 500Hz. The robot system has the abilities to adjust the strength level and to change the strategy based on the game situation. A system designer can easily adjust these abilities by setting several specified parameters. In this paper, first, the response control to adjust the strength level of the robot is explained. Secondly, the decision-making method using AHP (Analytic Hierarchy Process) is proposed. This ability enables the robot to switch the game plan. Thirdly, the catching motion for the air hockey robot is explained. Finally, we show the data of experiments and verify the effectiveness of these functions.

The goal of this study is to develop the technology of the robot to reach the goal of a human player for more than 80 percent. To achieve this, we use high speed manipulator, high speed image processing method, real-time and high-accuracy control system. Therefore, we need to measure a puck position accurately through the image processing, estimate the motion of the puck steady with particle filter, calculate the motion of the puck with the law of conservation of momentum. In this paper, first, the position of the puck is measured by removing noise from the image processing. Secondly, the motion of the puck is estimated by using recursive least square estimation, considering the reflection coefficient and rotation of the puck. Thirdly, the trajectory of the arm is generated so as to reach the goal. Then, we show the data of experiments and verify the effectiveness of the system.

In this paper, we explain a simple model and a deformation control of a linear flexible object with a high-speed robot arm. First, we propose the simple deformation model of a flexible rope using a high-speed robot motion. Then, we perform a theoretical analysis of the rope deformation model. And, we obtain a motion condition of the high-speed robot arm based on the theoretical analysis. Next, we show experimental results of the rope deformation control using the motion condition. Finally, we show the dynamic knotting of the flexible rope as an example of the rope deformation control.

The goal of this study is to develop a robot vision system which can recognize colors more precisely than human at high speed. To achieve this, we have developed a system which consists of a multi-spectral light source which can change its wavelength at high-speed and a high-speed monochrome camera. The system can recognize and track a target object at high speed. First, this system measures the spectral information of the target surface, and finds the features which are remarkable about spectral reflectance. Secondly, by changing the wavelength of illumination actively, the system tracks the target with the features of spectral reflectance at high speed. Experiment results, in which the system distinguishes a target from other objects which has very similar appearance in human eye, are shown.

Humans perform skillful and fast manipulation by utilizing the dynamics of the targets effectively. Visual sensation is the most important human sense for such manipulation. Juggling is one example of skillful and dynamic manipulation of human, and there are several previous researches about robotic juggling. However, there are few researches in which a human-like multi-fingered hand-arm is used. The purpose of this study is to achieve ball-juggling with a hand-arm, which has three general purpose fingers, and stereo vision. Image processing is executed at 500fps by high-speed vision system and GPU. The trajectory of the hand-arm is generated based on the estimated ball's dropping position and the moment, and the hand catches the ball. The juggling motion is achieved by repeating this cycle. As a result, the robot has succeeded in juggling one ball.

As for human, skillful manipulation is achieved by using the dynamics of the target effectively. Also as for robot manipulation, juggling is an example of skillful manipulation, its efficiency is improved by using unstable condition of the manipulated object. There are many previous researches about juggling. However, there are few researches, in which a multi fingered hand-arm like a human is used. In this paper, at first, a ball position is extracted by high-speed image-processing. Next the ball trajectory is predicted by RLS (Recursive least squares) approximation. At last, the experimental result is shown in which a ball is thrown by the hand-arm and its trajectory recognized by the high-speed vision.

The purpose of the study is to achieve high-speed throwing motion using the hand-arm. In order to achieve high-speed motion, an appropriate motion using the dynamic characteristic of the arm is needed. Most of human sports players can use kinetic energy efficiently based on dynamics of body. As a result, they achieve high-level performance. In this paper, sine waves are used to imitate human throwing motion. The advantages of the proposed algorithm is simple to set an objective throwing motion. The parameters are optimized based on the dynamics of the hand-arm. In addition, a follow-through algorithm is added using polynomial interpolation. Experimental results are shown.

The goal of this study is to develop the technology of high-speed human interactive robot in which the robot reads the opponent's intention and moves in response to the opponent's motion and human purpose expectation. To achieve this, we develop an air hockey system using a high speed manipulator, high speed image processing method and real-time control system. The manipulator can play the air hockey game based on a recursive trajectory generation using continuous images from high-speed vision. First, the position of the puck is measured by removing noise from the image sequences. Secondly, the motion of the puck is estimated by using recursive least square estimation. Thirdly, the trajectory of the arm is generated so as to defend the goal. Then, we show the data of experiments and verify the effectiveness of the system.

The purpose of this paper is to achieve skillful manipulation by a multi-fingered robot hand with fingertip revolution axes. First, the kinematic relationship between the fingertip revolution and the orientation of a grasped object is proposed. Secondly, based on the kinematics, a new regrasping method of a grasped object is proposed. This control method uses the rolling contacts between the fingertips and the object.

The purpose of this paper is dynamic manipulation of sheet flexible object. As one example of dynamic manipulations of sheet flexible object, dynamic folding of a cloth with two high-speed multifingered hands mounted on sliders system will be carried out. First, the dynamic folding by a human is analyzed in order to extract the necessary motion for achievement of this task. Second, a model of sheet flexible object will be proposed by extending the linear flexible object model (algebra equation) using high-speed motion. Third, the motion planning of the robot system will be performed by using the proposed model and the simulation result will be shown. Finally, the experiment with the trajectory obtained by the simulation will be executed.

In this paper, we focus on control of slip of paper. The goal is the position control of a paper by recognizing slip information. For this purpose, we use high-speed multifingered hand with CoY (center of pressure) sensor. CoP sensor can output the center of pressure at contact position, and it can also output the vibration information caused by slip. By analyzing the information with Wavelet transform, the hand is controlled so that the paper is set to the desired position. The results of the experiment are shown.

This paper proposes a teaching system for multi-fingered robot hand using kinetic information of a manipulated object under rolling contact in order to achieve the desired movement of a manipulated object easily and accurately. According to the kinetic information of manipulated object which was obtained by haptic device, the teaching system sets the contact position and calculates the trajectory of each joint angle by inverse kinematics function. To set the contact position appropriately, the system considers constraint condition under twist rolling contact. This paper describes the results of experiment using the developed teaching system, and the future works to be improved are discussed.

In this paper, a new zoom and focus mechanism for a high speed-vision is proposed. The mechanism consists of a varifocal lens and two AC servo motors. It is lightweight, and it has good performance to be controlled at high speed. We propose a visual servoing control for the zoom and focus mechanism. As the image features, the area of a target and SML (Sum of Modified Laplacian) are used. The relationships between these features and the joint angles of motors are measured to decide the parameters of control.

Multifingerd hand is capable of using tool and can increase its ability. With high-speed actuator and vision robot can do more advanced tool manipulation than human. The task multifingered hand catches a small ball in parabolic and free fall motion using tweezers is taken up for its demonstartion. Experimental results are also shown.

In order to achieve a skillful handling like a human hand, it is necessary that a hand has a capability to manipulate tools. One important problem of tool handling with a multifingered hand is that relative positions among a robot hand, a tool and a handled object change during the handling. For this reason, it is necessary to measure the relative positions in realtime, and to control a hand so as to cancel the changes. In our research, we have resolved this problem by using a high speed visual servoing and a passive joint model. Using the passive joint model, dexterous and flexible tool handling is achieved. In this paper, as an example, a tweezers is handled by a multifingered hand. A new method to control tools and some experiment results are shown.

In this paper high-speed manipulation system is developed, which consists of two manipulators and active visions. The design concept and the features of the system are described. In addition a task using non-contact state is taken up in order to achieve high-speed dynamic manipulation. Experimental results are also shown in which a manipulator throws a ball and the other manipulator hits the ball.

In this paper, we propose a new zoom and focus tracking system. First, a new zoom and focus mechanism for a high speed-vision is proposed. The mechanism consists of a varifocal lens and two DC blush motors. It is lightweight, and it has good performance to be controlled at high speed. Next we design a control system for a high-speed zoom and focus mechanism for a high-speed vision. In the control system, zoom is controlled by the area of a target, and focus is controlled based on SML(Sum of modified Laplacian).

In this paper we propose a novel display system that consists of a high-speed vision and high-speed projector. And this system can control a view of the projected image by controlling the original image acquired from high-speed vision. The projection image is generated based on the affine transformation estimated from present and desired positions of feature points on the camera image. By repeating these processes, a desired view of the projected image can be achieved on any deformable screen.