Increasingly, light and fast robots are in demand, and the flexibility of their links are no longer negligible. For such robots, which are often called fllexible arms, it is necessary to compensate for deformations and to supress vibrations caused by the elasticity of the links. In general, some mathematical models are required for the analysis or control of flexible arms. There have been many studies about planar motions of flexible arms, but not so many about 3D motions. In this paper, a 3D spring/mass model is proposed to describe an elastic link with a heavy lumped mass and a light beam in 3D motion. A flexible arm which has three joints and two elastic links is modeled in this way. The equation of motion and then the state equation are derived to describe the dynamics of the flexible arm under certain assumptions. It is shown that we can determine the end-of-arm position and orientation from the output of accelerometers installed at the tips of the elastic links. Experimental results demonstrate that a controller designed using optimal regulator theory can suppress the vibration significantly.
To enhance the ability of robot manipulators to be applied to assembly tasks in unstructured environments, intelligent control to extend the autonomy of the manipulator is urgently required. The concept of the intelligent control, however, has not been established although many approaches have been presented. This paper presents an approach to the implementation of artificial skill-artificially realized human physical skill to be controlled intelligently-that combines the framework of symbolic manipulation by Lisp and the framework of hybrid control of positions and forces that has been presented for the control of constrained motion. First, on the analogy of human brain that implements natural skill, the control system is designed hierarchically, of which the upper level consists of recognition and planning systems based on specially designed rules and the lower level consists of a hybrid controller written by C. An important part is the interface between the upper and lower levels, which is designed in order for the framework of the hybrid control to be matched with the upper level. The above design framework is applied to a special task: a peg-in-hole task, to yield a skill system for the task. Simulation results show that the system controls the task successfully.
A computer-aided design system of robot manipulators has been developed, which is one of the subsystems of“TOCARD (Total Computer-aided Robot Design) ”system. This system determines the detail of robot structure (not only arm dimensions but also joint sizes and machine elements such as gears and chain transmissions) of any type of robot to minimize its total weight under the constraints of deflection and natural frequency, selecting machine elements from catalog data. In this system“sensitivity”is introduced to reduce man power and make interactive design efficient. The objective function for optimization, which consists of total weight, deflection, and natural frequency, is obtained applying the penalty method to the optimal design problem of this system; the sensitivity of design parameter is defined as the differential of this objective function with respect to the parameter variation. Using this sensitivity, a designer can pick up the most sensitive parameters for optimization. Furthermore, arm cross sections are automatically converged to their optimal solution by the penalty method after the other design parameters come near to their optimal value in the above-mentioned interactive optimal design step. These interactive and automatic steps permit efficient optimal design of robot structures.
A method for multiple sound source localization is presented. It is based on the properties of the onsets of sound as the primary cue: (1) most of the onsets will be contributed by only one source. (2) the onsets will be contributed only by the direct sound. Histograms of temporal disparities for several frequency components and its integrated one are calculated to determine the correct temporal disparities. Experiments were carried out in an anechoic chamber and an echoic chamber. The recorded speeches were used as two sound sources. We could localize the two concurrent sound sourcess with reasonable accuracy.
A problem of the stereo vision is that measurement error occurs for curved objects. Boundaries of curved objects can be classified to two classes; discontinuity boundaries where two surfaces intersect physically and positions of discontinuity boundaries where a surface turns smoothly away from the viewer. The 3-D positions of disconinuity boundaries can be obtained by the stereo method, but measurement error occurs for the extremal boundaries. On the other hand, at an extremal boundary the surface orientation can be obtained by“shape from contour”method. Therefore, to get the 3-D information of curved objects it is necessary to classify boundaries to two types; however, few method has been proposed so far. In this paper, we propose a method to classify boundaries to two types by trinocular vision. When a curved object is viewed from two cameras, the intersecting point of two lines of sight is on the discontinuity but is not on the extremal boundary. This gap between the intersecting point and the boundary can be detected by the third camera: thus discriminating the extremal from the discontinuity boundary. We have made experiments on several types of cylinders and obtained satisfactory results.
In this paper, a position/force hybrid control of a robotic manipulator based on a neural network model is proposed with consideration of the dynamics of objects and the orientations of the robotic manipulator. This propsed system consists of standard PID controller, the gains of which are augmented and adjusted depending on objects and orientations of manipulators through a process of learning. The authors proposed a simular method previously for the force control of one degree-of-freedom manipulator. The proposed method shows the better performance than the coventional PID controller, yielding the wider range of aplications. This paper shows the similar structure of the controller via the neural network model applicable to the cases of the position/force hybrid control of multi degreesof-freedum manipulators. Simulations as well as experiments are carried out for the case of the two dimentional robotic manipulators. The results show the applicability and adaptability of the proposed method to the position/force hybrid control of manipulators.
Servo parameter tuning for industrial robots before delivery is very important, because the trajectory-tracking accuracy depends mainly on the tuning results. However, conventional trial and error tuning work is time-consuming. This paper describes a new servo parameter tuning system for industrial robots. This system is composed of a robot identification function and a servo design function. The identification function is based on an accurate frequency response analysis, called Multi-Decimation (MD) method. The servo design function is based on a stable method in the cut-off frequency region for calculating a two degrees of freedom (2 D. O. F.) PID digital servo parameters. In a SCARA robot used for tuning experiments, vibrational characteristics, caused by speed reduction mechanisms, extend over a wide frequency range. The proposed tuning system could be used to successfully identify the wide range of frequency response, using input-output data within only two seconds. Also, optimal servo parameters were obtained by the servo design method. Consequently, tuning time for a new SCARA robot has been shortened to one-tenth the previously required time. The system has been implemented in a portable laptop-computer. Therefore, it is easy to retune servo parameters for changing payloads, and to maintain robots in a customer's factories, without an instrument such as an FFT servo analyzer.
This paper describes a model-based robot system. This system was developed to replace and improve upon the conventional robot control scheme of teaching-by-showing and replay. The key concept of this system consists of the combination of manipulation skills with an environment model. The model provides the geometric structure and physical properties of the objects in the environment. An interactive geometric modeling system has been developed to generate accurate model through a good manmachine interface. The manipulation skills enable reliable task execution in the presence of unavoidable errors and uncertainties. Using the ETA-3 direct-drive manipulator, skills are implemented within the hybrid (position/velocity/torque) control scheme and are incorporated with sensing procedures to detect the state to be achieved, disassembly of a valve has been successfuly achieved by using this system.
We report a method for obstacle avoidance control against moving or standing obstacles for mobile robot. The speed of moving obstacles is uncertain, and their speed is expressed using fuzzy numbers. This paper presents an obstacle avoidance algorithm using fuzzy production system. Some results of computer simulation are shown to evaluate the effectiveness of the proposed method.