This paper presents a knowledge model of everyday objects for semantic grasp. This model is intended for extracting the grasp areas of everyday objects and approach directions for grasping when the 3D point cloud data and the intended purpose are given. Parts that make up everyday objects have functions related to their manipulation. We therefore represent everyday objects in terms of connected parts of functional units. This knowledge model describes the structure of everyday objects and information on their manipulation. The structure of an everyday object describes component parts of the object in terms of simple shape primitives to provide geometrical information and describes connections between parts with kinematic attributes. The information on the structure is used to map the manipulation knowledge onto the 3D point cloud data. The manipulation knowledge of the object includes the grasp areas and approach directions for the intended purpose. Fine grasps suitable for the intended task can be generated by performing a grasp planning with consideration for stable grasp and the kinematics of the robot in the grasp areas and approach directions.

We examine postural reaction and balance recovery patterns occurring when a standing upright human is subjected to a sudden disturbance within the frontal plane, with the aim of developing balance control strategies for humanoid robots. Five patterns are identified and related to the magnitude of the disturbance. Three of the patterns are modeled and implemented with a small humanoid robot HOAP-2. The models are based on inverted-pendulum and double-pendulum equations, in combination with variable stiffness/damping elements for ensuring appropriate reactions and balance recovery patterns. Supporting foot reaction is minimized within the reaction null space formulation. Special attention is paid to the transitions between the reaction patterns. The experimental data show that the models and the respective controllers can ensure smooth reaction control under both impact-force and continuous-force disturbances.

High-speed path tracking with a kinematically redundant manipulator mounted on a flexible base is addressed. Thereby, possible vibrations of the base are to be suppressed. In general, the presence of kinematic redundancy allows these two subtasks to be performed simultaneously. In practice, however, this can be done only within very limited areas of workspace, separated by singularity loci that change dynamically while the end-effector tracks the desired path. To avoid controller performance degradation in the neighborhood of such dynamic singularities, and to allow transitions between the distinct workspace areas through singularity boundaries, we propose here a new method for reactionless motion generation within a specified neighborhood of the singularity. In contrast to previous works, this method makes use of a nonzero coupling momentum which is conserved upon entering the neighborhood.

Our ultimate goal is introducing energy-efficient walking patterns to actual humanoid robots. It is known that limit cycle based walking methods, e.g. Passive Dynamic Walking, have such a desired property. Unfortunately, the application of the methods has been limited to simple planar biped models. In this paper, we propose a way of extending limit cycle based walking pattern generation toward a 7DOF 3D biped with ankles, knees, an upper body and with flat feet. This is achieved via first decoupling roll and pitch motions in the frontal/sagittal planes, and then, by designing a limit cycle based walking pattern in the sagittal plane for a planar 5DOF model with ankles, knees and torso. Robustness of the motion is ensured via a feedback control method based on mechanical energy. Then, the planar motion pattern is projected back into 3D space by incorporating dynamic components for gravity compensation and designing a proper trajectory for ankle roll motion. The performances of the walking pattern generator and the controller are confirmed via numerical simulations. The results are presented also as animated motion of a 7DOF 3D biped in the accompanying video.

We aim at experimental verification of the efficiency of natural-motion path tracking (i.e. tracking speed in proportion to the determinant of the Jacobian) in comparison to constant-speed path tracking. This is done first via simulations, with a simple planar manipulator and then with a six-DOF manipulator. From the results it becomes apparent that natural-motion path tracking outperforms constant-speed path tracking in terms of peak joint speed, peak joint torque and total mechanical power, ensuring thereby a higher average tracking speed. The results are also confirmed via experiments with a real six-DOF robotic limb.

This paper describes 3D biped walking generation and control based on Limit Cycle Walking. In our study, we use the simplest possible 3D biped model with three DOFs, incorporating roll and pitch motions in the frontal/sagittal planes, respectively. Our approach dynamically decouples these two motions, stabilizes pitch motion in the sagittal plane via the Limit Cycle Walking approach, introduces robustness for this motion using energy feedback control, and robustness for roll motion based on reference trajectory feedback tracking control. The roll reference trajectory is generated via analysis of the impact dynamics. Performance is verified via simulations.

In this paper, a method for displaying feeling of cutting by a micro-scissors type haptic device is described. Micro-scissors are a type of surgical instrument, which is frequently used in brain surgery under a microscope. A prototype of a micro-scissors type haptic device consists of three components: an interface, a drive member, and force sensors. The interface is designed to equip features of a pair of micro-scissors and has a blade spring on its end. The drive member is composed of two DC-motors and two crank-lever mechanisms as decelerators. The force sensors are composed of strain gauges with H-slits, and implemented into each handle of the interface. For the aid of applying the force sensors to feedback control of cutting resistance forces, compensation for influence of a blade spring on the force sensors is added. Evaluation experiment for basic performance of the device is carried out, and the device is proved to be able to generate the computed cutting resistance forces. Virtual cutting experiment by six subjects is carried out, and it is proved that the device is able to display feeling of cutting.

This paper describes a procedure of avoiding singularity for a redundantly-driven, dual-arm, master-slave robotic system. In cooperative operation, the procedure starts with a periodic check if one of the dual arms is close to a singular configuration by examining the manipulability of the arm. If it is close to a singularity, then the system stops its cooperative operation and control scheme switches to redundancy control. Then the operator manipulates the joint until the arm moves away from the singular configuration by inputing angular velocity to a redundant axis. The fact that the operator can select a new arm configuration to avoid singular configuration in teleoperation is the contribution of this research. After moving away from the singular configuration, the system can resume cooperative operation. Because this manual operation applies to a redundant axis only (in the null space of the Jacobian matrix), it does not affect the current end-effector pose and force status. Experimental examples are provided to demonstrate the proposed method.

In order to verify orbital operations of a dual-arm space robot on the ground, a hybrid simulator (hardware in the loop simulator) is developed. The hybrid simulator includes a 14-degrees-of-freedom (14-DOF) dual-arm robot and 9-DOF motion table. A hybrid simulator has a great advantage in simulating complicated collision with multiple contacts, because it is difficult for a numerical simulation to get reliable and accurate results of such complicated phenomena. In this paper, the system architecture and the motion planning for three motion tables are presented. Two experiments are performed to confirm the basic motion and simulate a free-flying dualarm space robot colliding with a floating object. The feasibility of robot operation is discussed from showing the position and force data obtained in the motion simulation.

A purpose of this paper is to expedite onerous initial procedures when a force-controlled parallel robot is taught task motions by the off-line task teaching/playback system which we have developed to realize the motion planning suitable for tasks and playback task motions dexterously. In this teaching system, the numerical teaching method has been employed to give detailed descriptions of the motion parameters, but it brings out a trade-off between fine-tuning and complication. Therefore, new devices and functions are added to the teaching system and multiple teaching modes are implemented for solving this problem. The main feature is the use of task teaching with human demonstration data which are obtained in real and virtual human motion measurement systems. The initial teaching data are generated by the demonstration data and the robot commands are converted into them with the control parameters. Typical task is demonstrated and we verify the effectiveness of the teaching modes experimentally

This paper purposes to implement fast and complicated tasks by a high-speed parallel robot, HEXA, for industrial applications and the others. Although this parallel robot does not use any force/torque sensors, it has sufficient capabilities to achieve various tasks by applying an integrated control of motion, force and compliance, which takes advantage of the back-driveability and friction compensation of the actuators. A key point is how simply and easily operators can bring out the performance. We develop an off-line task teaching system to realize the motion planning suitable for tasks. Some tasks are demonstrated by using this teaching and control system and we experimentally confirm the effectiveness of our system.