This paper presents a novel parallel processing manipulation system, developed to provide extensibility and capability of real-time response. The system has been organized based on multi-agent using many transputers. The advantages of the system are 1) capability to integrate manipulator controls from higher intelligent levels through lower servoing levels in which real-time response is strongly desired, and 2) flexibility to accumulate various methods and to utilize them any time without bad effects to the system. First, in this paper, an important issue to be considered for system design is introduced as a result of a classification of data processing from sensors to actuators. Next, guides to design the system are presented based on the issue. Third, system hardware and software are described in detail. Finally, a prototype system for direct drive manipulator, ETA 3s, is shown. The effectiveness of the system is demonstrated with two extended systems for different studies of adaptive control and pseudo contact position monitoring, developed with low cost.
A thin electrostatic linear actuator using plastic films has been developed. The film actuator consists of two sheets, a stator embedded with three-phase striped electrodes and a slider coated with a slightly conductive material. First, voltages are applied on the stator electrodes to induce electric charges on the slider. Then, pulse voltages are applied to generate propulsive force and have the slider traverse on the stator. The stator sheet has been fabricated by forming 240 [μm] pitched copper lines on a polyimide film and covering them with a polyimide film. As the slider sheet, a 12 [μm] PET film coated with mixture of polyurethane and carbon particles has been used. Total thickness and mass are 120 [μm] and 0.35 [g], respectively. This actuator produced 0.16 [N] propulsive force and 1.6 [mW] output power at 800 [V] driving voltage. The film actuator is expected to be used in variety of applications such as display devices, thin material conveyers, or powerful artificial muscles.
This report presents a general method toward discovering unknown skills which are required to achieve some dexterous tasks. We define a skill as a domain specific mapping from sensory input to control output and formulate skill discovery as a strong learning of such skills: strong learning assumes little a priori information of the task. Apart from the context of manipulation or assembly tasks, our definition of skill involves dynamic and reactive behavior. We approximate a skill function by a polynomial function and reduce skill discovery to a combinatorial optimization problem among the coefficients of polynomial terms. Our approach applies genetic algorithm (GA) to this problem because GAs require no a priori knowledge of the problem and are shown to be competent to solve combinatorial problems. Practical issues about how to enable GAs to make efficient search are also discussed. As a preliminary test case of the proposed framework, we consider a simulation setup to discover a skill to catch a ball flying in 3D space. A mobile robot on a plain is given elevation and azimuth angles and their time derivatives of the ball from the robot viewpoint and is commanded to catch balls projected with various initial velocities. The skill to be discovered is a mapping from this visual information to a 2D driving force vector. Simulation results show that the proposed method is successful in discovering the skill to catch balls and that the discovered skill is tough to some extent against such disturbances as the influence of the wind. Finally, the discovered skill is analyzed and found to be compatible with the strategy of a human ball catcher.
This paper describes improvements to the mechanism and control of an in-pipe magnetic-wheeled robot. The robot has dual magnetic-wheels and can turn in a T-shaped iron pipe. However, the conditions for the turn are extremely restricted, that is, the robot must move within a certain tiny area of the pipe. To ease the conditions, a mechanism which allows the robot to twist in the middle has been successfully adopted. Meanwhile, it is very important to control the robot's movements in the area of the straight pipe before the turn. To guide the robot in the area, feedback controlling the angles of steering based on the angles of the steering and rolling is efficient. Futhermore, a Kalman filter is used to estimate the angle of steering. Sensing the angle of the twist is also an efficient way to compensate for the limitation of the sensor of the angle of rolling to improve robustness. An experimental robot has been successfully fabricated to confirm the efficiency of these ideas.
This paper presents a method for optimal design of local communication area. Despite many studies of local communication for multiple mobile robots, communication area has not been designed based on mathematical analysis, but on only time-consuming simulations of many-robot communication. We therefore analyzed the efficiency of information transmission, and derived the optimal communication area which minimizes the information transmission time to multiple robots. This optimization mainly consists of two steps. First, we derive the“information transmission probability”for various task models. And next, as the evaluation function to minimize, the information transmission time is represented using the derived information transmission probability. The analytical results are also verified by computer simulations of many-robot communication.
A dexterous micro manipulation system is developed for application in assembling micro machine, manipulating cells, and micro surgery. We propose a concept of a two finger micro hand, then design and build a finger module based on parallel mechanisms. The module has 6 degrees of freedom with piezo electric devices as actuators. A calibration method based on least square error is proposed for translational motion in task coordinates fixed under a microscope. Coordinated motion is generated by both programming control and joystick teleoperation control. Basic characteristics, such as positioning accuracy, step response, and working space, are evaluated.
A design concept of TOMMS, TOshiba Modular Manipulator system, has been already proposed to achieve a modular manipulator system which can be assembled into any desired configuration to provide adaptability to tasks, using as few kinds and types of modules as possible, without special handling such as modification of control software. To realize the concept, we developed a constitution and configuration recognition method of the assembled manipulator using electric resistance which is simple, practical and reliable. Moreover, to actualize the system which can offer the best suitable manipulator constitution and configuration for the desired task, we developed a workability judgment method considering the degeneracy of degreeds of freedom (d. o. f.) of the manipulator and the conditions of the desired. These methods were applied to the trial system TOMMS-1 and their efficiency and practicality were confirmed.
It is obvious that visual servo systems have various applications for real time robot control. But most conventional vision systems have serious restriction on their performance, since those systems always use CCD cameras to acquire the images which transmitted by serial video signals. Therefore the sampling rate is limited by the video frame rate, and this restriction on the speed is quite insufficient to control of the robot. To solve this problem we have developed a massively parallel processing vision system called SPE-256 in which the photo-detectors and processing elements are directly connected. We have realized high speed visual feedback with 1ms cycle time. In this paper we describe our 1 ms visual feedback system and its performance in the application of high speed target tracking.
Measuring the force acting on a flexible wire by using a force/torque sensor is difficult when the flexural rigidity of the wire is small, because the wire buckles and the force acting on the wire is smaller than the force that can be measured by the force/torque sensor. This paper shows a method of estimating the force acting on a wire from its shape observed by stereo vision. The force is obtained continuously by using visual tracking. Consequently, we propose a strategy of inserting the wire into a hole on a wall by using the estimated force. Three experiments have been successfully carried out.
We propose a method to incrementally make a geometric model of objects/environments by stereo vision. The problems to solve here are not only how to extract 3D geometric data of a scene which a stereo vision system observes but also how to integrate partial and incomplete geometric data which are obtained from different views into a geometric model. We use a b-rep (boundary representation) for geometric models.
In this paper, we describe a method for generating an environmental map by cooperative observation among multiple mobile robots with omnidirectional visual sensors. Each robot with a conic projection image sensor COPIS can observe an omnidirectional view around the robot in real-time with use of a conic mirror. Under the assumption of the known motion of the robot, an environmental map of a scene is generated by monitoring azimuth change in the image. First, each robot communicates and exchanges sensory data information. Then, static objects, unknown moving objects and other cooperative robots are discriminated by evaluating relative directions of motion and estimated locations of vertical edges. Finally, a global environmental map is generated by matching all environmental maps.
The integration theory of reactive behaviors are to be discussed in this paper. A linear emerging model is adopted where the motion of a robot is represented as the weighted linear sum of reactive behaviors. The weights are defined as differentiable nonlinear functions of sensor signals and parameters. The functions can represent logical if-then rules as their extreme cases. The sensor space model is introduced to relate the sensors and the behaviors and to determine the parameters. We establish a learning method based on the sensor space model, where the parameters are systematically tuned through iteration of trials such that the sensor signals converge to the given teacher signals. A nonlinear dynamics in the sensor space model is also proposed to allow fluctuation for the future global search. The learning method is applied to the reactive grasp of a three-fingered robot hand. We integrate 48 kinds of sensor signals and 29 primitive behaviors. The experiments indicate that the emerging model allows us to use the semantics to initially program the nonlinear functions for the weights. The learning experiments successfully illustrate the usefulness of the proposed learning method.
Derivation of an equation of motion of closed loop mechanisms by the motor algebra is presented in this paper. The equation can be given by first determining the velocities and accelerations of the links, next deriving the equilibrium equations of the forces and moments on links, and then appling conditions of the virtual work of joint motions to the equilibrium equations. Simple description of the derivation method are shown by using the motor algebra. In order to reduce the computational cost for the dynamic analysis, it is essential to utilizing the geometrical specialities in a mechanism to the derivation of an equation of motion. It is shown, by two examples of the derivation for a translational table mechanism and a parallel mechanism, that the expression of derivation by the motor algebra makes their utilization easier.
We have been developing a teleoperation type robotic system for hot line maintenance. In the system, a dualarm robot is operated by manual control and automatic control. The range information of objects is very useful for the automatic operation. We have developed a range sensor system in which a range sensor measures the distances to an object and the robot manipulates the object based on the range data. This paper describes the range sensor system. The range sensor consists of a spot laser source and two CCD cameras, and computes ranges using triangulation method. In order to reduce the influence of sunlight, the laser emission synchronizes with the frame signal of the camera. A differential images are obtained by subtraction of two sequential raw images and a minimum image is computed from the differential images. The laser spot on an object is detected in the minimum image. Calibration is essential for a sensor-based robot. This paper also mentions the calibration method, and lastly the application examples for robot manipulation and robot base approach are presented.
This paper presents a real-time visual tracking system that can follow a moving object by controlling the pan and tilt of a camera based on the object position and speed obtained by video-rate image processing. The system consists of a camera, an image processing system, a pan/tilt platform and a motor driver. The developed platform can control the pan and tilt angle of a camera with accuracy of 0.002 [deg] and speed up to 50 [deg/s] . We have developed a tracking algorithm using both position and velocity of a moving object. The system integrates them to capture a moving object in the center of camera view. The algorithm, implemented on a general purpose image processor, can detect a moving object within 33 [ms] . As an experimental result, the system can track a moving object (walking person) in real-time.