In this paper, we propose the application of DSP (Digital Signal Processor) to real time computation of dynamic control for robot manipulators in low cost. There are some applications of DSP to robot control so far, but these applications are not to main computational part in robot control such as kinematics or dynamics.
Generally, many of DSP can not operate in floating point, but in fixed point. Therefore, it is usual to scale each physical value of each unit in order to obtain accurate fixed point expression. However, this scaling method is not sufficient in the case of the precise computation for robot manipulators, because the absolute physical values at each link may change in very wide range through every link while a manipulator is moving. Therefore, there is a possibility that computation result includes large error. In this paper, we propose a new scaling method regarding the structure of recursive computation for robot manipulators in order to avoid large computational error.
Then, we constructed actual control system by using a 16 bit micro computer as a host computer and TMS 32010 as a DSP. Lastly, we confirmed that DSP can compute the unified computation of kinematics and dymamics for general 6 d. o. f. manipulators in 3.74 msec though 16 bit micro computer requires 68.9 msec for the same computation. It is also shown that the total computation of dynamic control including trajectory calculation and feed back calculation for general 6 d. o. f. manipulators can be performed in 14.71 msec by using this DSP/microcomputer system and it takes more than 80 msec for the same kind of computation when only a 16 bit microcomputer was used.

View full abstract

It is useful for control of robot manipulators and for task planning if we have a quantitative measure of manipulating ability in positioning and orienting the endeffectors. Its this paper, a method for estimating the ability of robot manipulators under the environment with obstacles is proposed from the viewpoint of kinematics.
The obstacles which are close to a manipulator bring some restrictions to the motion of the manipulator. In this paper, we discuss two cases; the one is the case that the joint velocity vector must satisfy the equality constraint for the purpose of avoiding the obstacle, and the other is the case that the approaching speed to the obstacle is restricted. For each case, we derive a manipulability ellipsoid and a measure of manipulability. Lastly, as an example, an experiment and simulation results using an eight degree of freedom manipulator are shown. In this experiment, the endeffector is inserted into an inside space through a narrow slit, and the joint angle velocities are controlled so as to keep the proposed manipulability large.

View full abstract

The complexity of teaching task to conventional arc-welding robot by a human operator depends on the shape of an object to be welded (work), and applicability of the teaching data is restricted to the target robot only. CAD information is not fully utilized in such a method, therefore it is difficult to build a flexible production system using robots separately.
In this paper the authors propose a compact computer aided welding system which consists of a personal-computer and a conventional teaching-playback arc welding robot, and indicate the concept of off-line programming by a simple robot language. In the source program of the weld, places to be welded are indicated on the work model without considering both type of robot and location of the work.
In the system, weld information processing was divided into four levels ; task level, tool level, robot level and machine code level. At each level, necessary information is obtained from external files. In task level, source program of the weld is parsed and weld parts are extracted from the work model. In tool level, the motion of weld torch is generated on the work model at each weld line and between weld lines referring to both of work and torch models. In this level, the motion data is independent from the robot. In robot level, each angle of robot arms, which realize required torch angle, is calculated considering the movable range of the robot. In machine code level, angle data of robot arms and weld conditions are combined together and converted into robot codes, which are transmitted to the robot controller through a communication line.
These solutions are confirmed by images of torch or robot on display monitor at each level. The property of the executable robot code generated by this method was confirmed by actual motion of the robot.

View full abstract

This report describes a scanning type suction cup (SSC) specially developed as a suction cup for a wall climbing robot.
Although a wall climbing robot with vacuum type suction cup can move in spite of variation of wall materials and small unevennes of wall surface, it cannot walk essentially on the wall that contains cracks and gaps.
To cope with this problem, we proposed SSC that consists of several suction cup units. The SSC is operated by switching the ports between the vacuum pump and the suction cup units. The system makes use of the fact that the vacuuming time constant is much smaller than the time constant of pressure raising due to leak. This SSC has the following advantages
(1) Even though some of the suction cup units in SSC lie on cracks or gaps, the suction force can be generated by the remaining suction cup units.
(2) As the ports between the vacuum pump and the suction cup units on crack or gap are closed most of the time of a period, the energy loss for such suction cup units is very small.
(3) As a small size vacuum pump with light weight can be used, SSC is effectively used for an energy self-containing robot.
The basic characteristics is made clear through simulation and some-basic experiments using a specially designed SSC.

View full abstract