Stable graspings have been reported so far, however the prehension is treated in such objects that have simplicity in shapes like the building blocks of circular or triangular shaped. Therefore, it is not clear about how an complex-shaped object should be grasped. This made it difficult to determine support forces of the fingers uniquely, in general. We define a weight center angle with which the force directing the weight center crosses the normal line on the object, and we think the angle is the important factor to realize the stable grasping in the actual environment. Then, we give a clear definition of the stable grasping of a plain object, by making the distribution of a grasp force or a press force smooth around the object. We propose a calculation method to determine the rectangular components of the support force uniquely, at the designated points on the object. The result of the calculation shows the method is useful to solve the support force in controlling multijointed fingers.
This article, modeling a mobile robot as a point, discusses on a problem of trajectory planning in transient environment. When a robot moves over a plane, every stationary environment is representable as some subspace of two-dimensional Euclidean space while every transient environment, as some subspace of three-dimensional Euclidean space obtained by adding the time axis to the two-dimensional space. Given a transient environment E and a start point B and a goal point G of the robot, by applying a paint procedure supported by graphics library in computer, examine an area DB being accessible from B and another area DG being accessible to G. Next, select an arbitrary point M just on the boundary between DB and DG. Then, the original problem is decomposed into “a problem of trajectory planning from B to M in environment DB” and “another problem of trajectory planning from M to G in environment DG.” By repeating this decomposition, each problem becomes primitive problem joining some adjacent two points in small environment. The trajectory joining the original B and G is a sequence of many short trajectories solved thus.
A new 6-axial force sensor using the optical measuring technique is proposed. The structure of the sensor is based on the unit which has a small light source being set face to face with a photo sensor of quarter-splitting type to measure their minute displacements in two directions each other. Three sets of the said unit are held by an elastic frame. The new sensor, in comparison with conventional strain-gauge-based device, is more compact, light in weight and low in cost. The amplifier can be integrated in the sensor itself. Compliance (approximately 100μm/full) is measured a little larger than conventional devices, but, the range can be decreased by further contrivance. The accuracy of the measurement was better than conventional devices by using newly introduced calibration method. Conventional calibration method for a 6-axial force sensor solely depended on a stiffness matrix induced under the state in which single force/moment was applied. In actual operation, however, non-linear interference occurs among plural force inputs. The newly introduced signal-processing technique has the features to make calibration under the generalized conditions. An experimental setup in which 6-axial force sensor could be simultaneously inputted with plural axial forces was produced and the proposed calibration method was demonstrated valid.
A new type of robot whose joints are supported by magnetic bearing mechanism is proposed. This robot joints are supported without mechanical contacts. It can avoid dust generation and oil lubrication which often become problems when robots are used in clean rooms, in vacuum chambers, or in space. In addition to the cleanliness and freedom of maintenance, the robot has a number of advantages as the freedom of friction, and the functions as micromanipulation, force control, force sensing, active vibration control, and so on. A prototype manipulator which has magnetic bearing joints is made. The experimental results on the magnetic bearing joint show the possibility of the proposed functions. And the levitation experiment of the prototype robot link shows that the robot joints can be supported by magnetic bearing mechanism and shows the feasibility of the a robot removing all mechanical contacts.
For the control of a dynamic mobile robot, the attitude in the gravity space is an important state of the robot. Usually, the attitude is difficult to detect by simply using the signals from sensors. For example an external sensor contacting the ground gets disturbances from the roughness of the ground; the integration of a gyroscope signal has the problem of drift; the inclination sensor does not indicate the direction of gravity when acceleration exists. To solve these problems, the authors propose a control method in which the attitude of a mobile robot is estimated by an observer considering the robot dynamics and using only the information obtained by internal sensors. The authors applied this method to a wheeled inverted pendulum as an example of a dynamic mobile robot. The estimation of the attitude was made with good accuracy using the signals from the rate gyroscope and the motor encoder, and the control of stable running of the pendulum on a flat level plane worked successfully. The authors also realized the running control of the pendulum on an unknown rough road using the estimation of the slope gradient by the observer. Thus, the effectiveness of the proposed method was demonstrated experimentally.
This paper proposes“skill”of manipulator to execute complex tasks. Skills are essential and elemental motions obtained by analysis of motions needed to execute target tasks. Since the skills are made to be adaptable to the task environment using force control, sensor feedback and so on, the elemental sub-tasks can be reliably accomplished with the corresponding skills. Highter level system can program or plan to execute a complex task by combining skills without knowing complex control of them. In this paper, considering assembly tasks as realization of target contact states, we analyze contact states appearing in polygonal assembly tasks and skills for the tasks are obtained by the analysis. Some of them have been realized using DD-manipulator: ETA3, and successfully applied to insertion of square pole.
In the past several years, various continuous-time robust control methods have been proposed for robot manipulators. Most of them consist of linearization by state feedback and robust compensation by switching gain. In the digital implementation of these methods, the computation time required for linearization should be taken into consideration. On the other hand, robust compensation should be implemented at a short sampling period to achieve the specified tracking error bound. However, there is little consideration of these points in the previous researches. Consequently, the systems often lead to chattering and the specified tracking error bound can't be achieved. In this paper we propose a scheme for a hierarchical robust control system for robot manipulators which is composed of an upper level loop and a lower level loop. In the upper level loop, the input for linearizing compensation, the desired trajectory and the switching gain are computed at a low sampling frequency. In the lower level loop, the switching input are generated at a high sampling frequency. This scheme will make the computation for robust compensation very fast. The performance of this hierarchical system is analyzed in consideration of the sampling period of an upper level loop and the modeling error.