Tendon-driven mechanisms have been used mainly in articulated-finger hands because of the small size and the light weight. Recently, for the necessity of forcecontrolled robot arms, several researchers developed tendon driven robot arms. In the case, since robot arms are much heavier and longer than robot hand fingers, tendons increase in the length significantly and also must run through very complicated paths. Therefore usual N+1 type or 2N type tendon-driven mechanisms are not enough to move robot arms and a number of redundant tendons are used to divide the total load into each of tendons and reduce the individual burden. For example, due to the gravity force, if the load in the clockwise direction is different significantly from the one in the counter-clockwise direction, we can use different number of tendons for each direction, respectively. There are additional advantages in a redundant tendon mechanism. For example, we can move the robot arm with the other tendons when some tendons have been broken and we can also adjust the mechanical joint impedance using non-linear elasticity of tendons. In this paper, several basic issues on the joint compliance adjustability for a robot arm controlled with redundant tendons are discussed. Namely, local maximum sets of stiffness elements which are adjustable simultaneously, and conditions under which the joint impedance can be adjusted using non-linear compliance tendons are discussed.
A real-time action search for an autonomous mobile robot using Evolution Strategy (ES) is proposed. By searching optimal actions for facing situations in real time, the robots are relieved from the problems of categorization of the situations and a-priori-definition of rules. The basic idea of the real time action search is to balance computational amount with exactness of solutions. In order to realize this trade-off, a two-stage evaluation technique is introduced under a framework of Anytime Algorithm. The prediction and action watch dog systems are also used. Then, the feasibility is shown through the experimental results.
We developed an active forceps using shape memory alloy pipes for laparoscopic surgery. The temperature of Ti-Ni shape memory alloy is controlled by electric heating, so that the response is improved. This paper describes a design and characteristics of the active forceps.
Timely input sometimes can control periodic movement of mechanical systems. In the swing which is the parametric excitation system, when the center of gravity is moved in double cycle, the swing's amplitude increases, and it decreases when this phase is moved π [rad] . But it is difficult to control a swing in the wide range, because the cycle of the swing depends on its amplitude. This paper shows that the van der Pol's equation inputted the swing's angular velocity can compose the frequency observer, and proposed the new control form of a swing that the center of gravity is moved by nonlinear feedback which consists of equation's output.
The main purpose of this research is to study and develop a robust sonar system for mobile robots. In this way, some researchers have made sonar systems which calculate the cross-correlation between the transmitted signal and the received one. However, as the number of patterns of transmitted signals are practically limited, there would be possible erroneous measurements by sonar sensors, especially when the number of the sensors is increased. In this paper, we theoretically estimate the probability of erroneous measurement by the sonar sensors, which depends on the number of the sensors and the form of the transmitted signal. Then, we designed the shape of the emitted signals so that the probability of having an erroneous measurement is in a defined range.
We propose a new redundancy control method for robot manipulators based on visual servoing, which enables obstacles to be avoided while positioning an end-effector in its target position. The control inputs to the manipulator are calculated both from the image features of the end-effector, other parts of the manipulator and obstacles, that are obtained from a stereo vision system, and from the known kinematic manipulator model. Moreover, our method enables the manipulator to avoid singular configurations simultaneously while taking joints limits of the manipulator into consideration. However, it is not easy to obtain all necessary image features of the manipulator by image processing in a real workspace. We estimate image features that cannot be obtained by image processing or which are hidden by some obstacles. Necessary image features of the manipulator are estimated from calibrated camera parameters and the known kinematic model of the manipulator for obstacle avoidance. Experimental and simulation results reveal the validity and effectiveness of our proposed redundancy control method for avoiding obstacles and singular configurations.
In a telerobotic task execution, a human operator and a telerobot should hold common knowledge about the task environment. We propose intelligent monitoring to let the telerobot offer appropriate information to the human operator to cooperate the task execution. The intelligent monitoring provides a function which selects important scenes to help operator through monitoring camera. We propose snapshot function as an extension of the intelligent monitoring to apply it for super-long-telerobotics under constraints of communication such as time delay and small capacity. The snapshot function selects important still images from within the monitoring images. We constructed a telerobotic testbed which includes a connection through the international ISDN and typical space structure and conducted trans-Pacific teleoperation experiments. Experimental results showed the intelligent monitoring and snapshot function were effective to the cooperation between human operator and remote robot.
Gesture recognition using images is recently studied as a technology of human friendly man-machine interface. In this paper, recognition of 3D gesture by using range images is discussed. Two methods for recognizing head gestures from a sequence of range images are proposed. One uses differential range images for identifying ‘Yes’ motion, and the other detects and tracks nose position for identifying ‘Yes’ and ‘No’ motion. Experiments to recognize real person's gestures using a range sensor system are performed to show the effectiveness of the presented methods.
When an object is grasped by a multifingered hand with power grasp, there are an infinite number of possible power grasp forms, provided by combinations of the hand shape, the contact positions between fingers and the object, and the joint torques of the hand. To plan a power grasp, it is necessary to know, by some prior evaluation, which form is most suitable among these forms. This paper proposes a technique for automatically obtaining optimal power grasp. By analyzing the degree of freedom of the force space caused by frictional contacts between fingers and object, the conditions and mechanical properties of power grasp are described. Also, the region of feasible joint torques for power grasp is derived. Then, an optimal power grasp form is defined and its determination procedure is established. Lastly, numerical examples are performed to verify the effectiveness of the proposed approach.
This paper discusses the scale-dependent grasp. Suppose that an object is initially placed on a table without touching by human hand and, then he (or she) finally achieves an enveloping grasp after an appropriate approach phase. Under such initial and final conditions, human unconsciously changes the grasp strategy according to the size of object, even though they have similar geometry. We call the grasp planning the scale-dependent grasp. Focusing on column objects, we first classified the grasp patterns into a couple of grasps and extracted the essential motions so that we can apply the scale-dependent grasp to multi-fingered robot hands. We found that the grasp patterns should be also changed according to the surface friction and the geometry of cross section of object in addition to the scale. The basic strategies were verified by experiments. Finally, we considered how each strategy should be switched from one to another.
Space robots performing precise servicing tasks in unmanned facilities will be indispensable in space activity of the 21st century. This paper presents the development of a precise telerobotic system which is the world's first precise extravehicular robot aboard the satellite, ETS-VII. The robot has the features of skill and autonomy for handling various parts in space through a three-finger multisensory hand. Firstly, we discuss the requirements for in-orbit precise servicing by a robot and system constraints to integrate into the satellite, and present the total system design for space telerobotic experiment. Secondly, we describe the design of robot mechatronics adapt to space use. The launch lock mechanisms and reconfiguring method of the robot in orbit are also described. Thirdly, an architecture and design of the onboard control computer system to manage various tasks flexibly and safely are presented. Finally, the multisensory based control for precise task execution and development result of the flight model are presented. The system was launched, and initial checkout reveals the system works well in-orbit.
This paper proposes a magnetic measurement principle to develop a balance sensor. Basically, the sensor is composed of a spherical vessel sealing certain amount of liquid, a mushroom-shaped float and a permanent magnet. The circular plated magnet is fixed on the root of the float stem. According as the liquid moves in the vessel, the float changes the position of the magnet. And the Halleffect devices located around the vessel sense the change of balance with the aid of a signal processor. The density of the Hall-effect devices and the algorithm for determining the direction of the resultant acceleration are considered. Also, the results of experiments for magnet's position and the angle of balance are shown. The proposed method is effective in detecting all directions of the resultant acceleration of motion and gravity with uniform resolution, since the sensor has no rotation axes.
Animals change their gait patterns with environmental conditions. The adaptive changes of gait patterns can be observed in the experiment where a decerebrate cat was forced to walk on the treadmill. This experiment suggests that cats should memorize the locomotion pattern and adjust it in adaptation process. From this point of view, we control the quadruped walking robot to behave adaptively on the treadmill like this cat. For the control and adaptation law, we take the decentralized approach: each four limb is regarded as subsystem, and pattern generation and adaptation is achieved with only the local interaction among four subsystems. Such a system structure is supposed to be important on the adaptability and flexibility.