In this paper, we propose a new behavior-based architecture with three-grouped (reflexive, purposive and adaptive) agents that realizes both efficiency in attaining the mission of the robot and robustness against various kinds of failures that may occur in a dynamic environment. Grouped agents are directly connected to the special agent (Motion Executor) which determines the motion command of the robot. The reflexive-level group consists of agents with contact, infrared and ultrasonic sensors which maintain the minimal safety of the robot. The role of the purposive-level group is to efficiently navigate the robot to the final goal. The adaptive-level group stands by to recover from failure in the purposive-level group or deadlock situations. Individual agent-arbitration mean is prepared for each group in the Motion Executor. Experimental results show the effectiveness of the architecture.
There are several methods to measure precisely the fine positional and orientational variations of the end-point of a robot manipulator. The cube method which is useful for measuring the repeatability, joint compliance, and end-point impedance of a manipulator, etc. is a well-known one. But, there are two problems in applying the cube method. The one is what kinds of displacement sensors (or transducers) are appropriate for the method, and the other is how to certify the precision of the assembled measurement system using the cube method. In this paper, we introduce the laser sensor type fine deflection measurement system for a robot manipulator we've developed and the precision certification method with 3-axis turn table we've devised.
When a force is applied at the endpoint of a manipulator arm, the endpoint will deflect by an amount which depends on the stiffness of the arm and the force applied. There are several sources, for example, arm links, transmissions, reducers, and servo drive systems that produce deflections or vibrations of the industrial manipulator arm. We've investigated the measuring methods of the equivalent joint stiffness of a industrial robots with serial mechanisms. In this paper, two measuring methods are proposed. One is simultaneous measurement method, and the other individual method. According to the experimental conditions, the former is devided into two methods further, servoing state measurement, and mechanically fixed state measurement. The experimental results for the typical industrial robot, PUMA560, show that we can regard the equivalent stiffness of each joint to be linear.
This paper reports the result of analysis for the servo system which includes the disturbance and velocity observer. Recently, the improvement of control theory brings us much useful knowledge. For example, we can get a compensator based on Youla parametrization if only we make use of an identity observer and a state feedback system. And it is known that a term of free parameter in this compensator is equivalent to the disturbance estimation value. On the other hand, I have presented a minimum order observer which estimated disturbance and driving velocity on each axis of robot manipulator. And it has become known that the control system using the observer has robust performance in generally. Then this paper proves that the control system is equivalent to a Youla's compensator when a term of free parameter is fitted to a condition to compose type-1 servo system. Next, when we make use of this minimum order observer, we can compose a two degrees of freedom robust servo system. Finally, this paper presents that there is a restriction on the pole allocations of the control system in practice.
This paper proposes a unified coordinated motion control algorithm of manipulators, which is applicable to both manipulation of an object and a parts-mating task. First, we consider the manipulation of an object by impedance controlled manipulators ; the impedance of each arm is designed so that the apparent impedance of the manipulated object is specified. Then we consider the parts-mating problem using the same control algorithm that is proposed for the manipulation of a single object. We propose a method to design the impedance of each arm so that the relative motion of two manipulators has the dynamics appropriate for the assembly of two parts. The experimental results using two industrial robots will illustrate the validity of the proposed control algorithm.
We propose a novel active vision system based on structural tracking. In this system, the position and the attitude of an object fixed to the camera coordinate system is kept constant by active manipulation of CCD camera. A high speed recognition method is proposed based on matching real image with image generated from object model.“Recognizability” is also proposed as the index of easiness of recognition. This is the original attempt to answer the question of “What are the best camera position and attitude to watch the object?” The proposed method was implemented by using two DSP's and the life-size 6-axis robot manipulator. One DSP was used for image processing and the other for robot control. The proposed active vision system showed its efficacy with laboratory experiments.
Almost all real systems are nonlinear and their mathematical models cannot be obtained exactly. In particular, it is very difficult to design control systems for nonlinear multi-input and multi-output systems such as robot manipulators because of their nonlinear interactions. In recent years, control methods of exact decoupling and exact linearization have been widely investigated by using differential geometric approach. But they have problems; they need exact models of systems and controllers become very complicated. In this paper, we firstly propose a method of decoupling control for general class of nonlinear systems by disturbance observer. The proposed method can decouple nonlinear systems not exactly but approximately, but controllers become very simple, and it is applicable even if exact models of systems are not available. Secondly, we propose a method of decoupling of robot manipulators by applying the proposed method. There have been a lot of studies on decoupling control of robot manipulators based on disturbance observers. But most of them realize decoupling of robots only in joint coordinate space. On the other hand, the proposed method can easily realize decoupling of robots in the task coordinate. Finally, in order to illustrate performance of the proposed method, experimental results by using a 3-degree-of-freedom series DD robot are shown.
Recently, a strong demands of developing autonomous decentralized system has been arisen since systems have been increasing in their scale and complexity. On the other hand, biological systems can be said the ultimate decentral-ized systems, and are expected to provide feasible ideas to engineering fields. Among them immune systems work as on-line fault diagnosis systems by constructing self-nonself recognition networks. In this study, we try to apply this immunological self-nonself recognition networks to a gait acquisition of a 6-legged walking robot as a practical example.
A systematic approach to the design of admittance matrix parameters in damping control is presented. Manipulative operations such as part-mating and grasping must be achieved despite positioning errors of manipulated objects. Applying damping control, a manipulator has a capability of coping with the positioning errors by modifying its velocity according to the reaction forces acting on it. An admittance matrix, which characterizes the actual manipulator velocity, must be determined so that the operation can be performed successfully. It is, however, difficult for human operators to determine an admittance matrix adequately for a various kind of operations. In this paper, we will develop an analytical method to derive admittance matrix elements for part-mating operations. First, the process of part-mating operations is analyzed with regard to how the workpieces contact each other. Kinematic and static property at each contact state is also analyzed. Second, we formulate the condition for successful part-mating operations based on the theory of polyhedral convex cones. Then, we develop a systematic method to compute admittance matrices that can eliminate positioning errors. Finally, an experimental result is shown to demonstrate the validity of the proposed method.
This paper proposes the scaled bilateral control system which amplifies the sensation of force, length, and dynamics in tele-manipulation. The bilateral control should be utilized in the teleoperation where the physical scales between the master and slave environments are different. Scaling rules are applied to design the bilateral system with an extension of the impedance control type bilateral control system. The response of amplified sensation and the contact stability of the system are discussed. Also, the proposed system is verified with the experiments with a pair of linear sliders.